A History of the Japanese Language

h \) -o ~ h ' \ J \ I ~ ~\ ( h)) ;, t ~ 1; .\ ) -I tl : l l t A History of the / .., 1r ~ JAPANESE Lan...

Author: Bjarke Frellesvig

376 downloads 2579 Views 21MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

h \) -o ~ h '

\

J \ I ~

~\

(

h)) ;,

t

~

1;

.\ )

-I

tl

:

l

l t A History of the

/ ..,

1r

~

JAPANESE Language STARK E F RELLES VI G CA" u It li>G L

This page intentionally left blank

A History of the Japanese Language

Bjarke Frellesvig describes the development of the Japanese language from its recorded beginnings until the present day as reflected by the written sources and historical record. Beginning with a description of the oldest attested stage of the language, Old Japanese (approximately the eighth century AD), and then tracing the changes which occurred through the Early Middle Japanese (800-1200), Late Middle Japanese (1200-1600) and the Modern Japanese (1600 onwards) periods, a complete internal history of the language is examined and discussed. This account provides a comprehensive study of how the Japanese language has developed and adapted, providing a much-needed resource for scholars. A History of the Japanese Language is invaluable to all those interested in the Japanese language and also students of language change generally. Bjarke Frellesvig is Professor of Japanese Linguistics, Director of the Research Centre for Japanese Language and Linguistics at the University of Oxford, and a Fellow of Hertford College, Oxford. He is the author ofA Case Study in Diachronic P hono/ogy: The Japanese Onhin Sound Changes (1995) and the co-editor ofJapan and Korea: Contemporary Studies ( 1997), Current Issues in the History and Structure ofJapanese (2007) and Proto-Japanese: Issues and Prospects (2008).

A History of the Japanese Language Bjarke Frellesvig

CAMBRIDGE UNIVERSITY PRESS

Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo, Delhi, Dubai, Tokyo Cambridge University Press The Edinburgh Building, Cambridge CB2 8RU, UK Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9780521653206 © Bjarke Frellesvig 2010

This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2010 Printed in the United Kingdom at the University Press, Cambridge A catalogue record for this publication is available from the British Library ISBN

978-0-521-65320-6 Hardback

Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or third-party internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.

For my children and my wife

Contents

List of tables, maps and figures Acknowledgements List of abbreviations Introduction Part I

XX XXl

Old Japanese Early writing in Japan and Old Japanese sources 1.1

1.2

2

page xvn

Writing Introduction ofwriting in Japan 1.1.1 Writing in Japanese 1.1.2 Logographic versus phonographic writing 1.1.2.1 Adaptation of Chinese script 1.1.2.2 Logographic writing of Japanese 1.1.2.3 1.1.2.4 Phonographic writing of Japanese Man')X}gana 1.1.2.5 1.1.2.6 Senm)X}-gaki Problems of decipherment 1.1.3 Polyvalence and equivalence 1.1.3.1 1.1.4 Reading tradition Rebus writing 1.1.5 Examples 1.1.6 Sources Japanese words in foreign sources 1.2.1 Early inscriptions, wooden tablets and archival records 1.2.2 1.2.3 Eighth-century texts Poetry 1.2.3.1 1.2.3.2 Prose Others 1.2.3.3

Phonology 2.1

KO-rui and otsu-rui syllables Co, versus Co, 2.1.1 Phonetic reconstruction and phonemic interpretation 2.1.2 Sound values 2.1.3

11 11 11 12 12 12 13 14 14 16 17 17 18 18 18 20 20 21 22 23 24 25

26 26 30 30 31

vii

vm

Contents

2.2

2.3 2.4 2.5 2.6

2.7

3

2.1.4 Phonemic interpretation 2.1.5 Neutralization Consonants 2.2.1 Obstruents 2.2.2 Tenues versus mediae; medial voicing and prenasalization 2.2.3 Non-sibilant versus sibilant obstruents Other allophonic variation Phonetic transcription of a text Syllable and word structure Morphophonemics 2.6.1 Vowel deletion 2.6.2 Rendaku Proto-Japanese 2.7.1 Consonants 2.7.1.1 Secondary origin of Old Japanese mediae (!b, d, g, z/) 2.7.1.2 Distribution of Old Japanese mediae and liquid 2.7.1.3 Proto-Japanese syllable final nasals 2.7.1.4 Proto-Japanese glides 2.7.2 Vowels and diphthongs 2.7.2.1 Arisaka's Law; distribution of primary vowels 2.7.2.2 Secondary vowels and diphthongs: OJ /-wi, -el 2.7.2.3 Mid vowel raising 2.7.2.4 Lexical distribution of /Cwo, Cye, Ce, and Cwi/ 2.7.2.5 Changes between proto-Japanese and Old Japanese

Grammar 3.1

Verbs 3.1.1 3.1.2 3.1.3

3.1.4

31 33 34 34 34 36 38 38 39 39 39 40 41 42 42 43 43 43 44 44 44 47 49 49

51 Derivatives Auxiliary verbs Inflected verb forms: obligatory categories 3 .1.3 .1 Finite verb forms 3 .1.3 .2 Non-finite verb forms 3.1.3.3 Nominal Auxiliaries: optional categories 3.1.4.1 Formation on lexical verbs Inflected forms 3 .1.4 .2 3 .1.4 .3 Respect 3 .1.4 .4 Voice; causative and passive 3.1.4.5 Aspect, tense, negation, and mood 3.1.4.5.1 Aspect and negation 3.1.4.5.2 Tense and mood 3.1.4.5.3 Combination 3.1.4.6 Perfective 3.1.4.6.1 Functions Distribution of the variants -(i)te and 3.1.4.6.2 -(i)n3.1.4.7 Stative 3.1.4.7.1 Function Morphological stative auxiliary: -yer3.1.4.7.2 Periphrastic stative: -(i)te ar-, -(i)tar3.1.4.7.3 3.1.4.7.4 Analytic progressive

51 52 52 53 53 56 57 58 59 59 62 63 64 65 65 65 66 66 67 68 68 68 69 69

Contents

lX

3.1.4.8

3.2

3.3

3.4

Negative 3.1.4.8.1 3.1.4.8.2 3.1.4.8.3

Analytic forms Negative rhetorical questions Reformation of the paradigm of the negative 3.1.4.9 Simple and modal past 3 .1.4.9 .1 Functions 3.1.4.9.2 'Perfect' 3.1.4.9.3 Speaker commitment 3.1.4.10 Conjectural and subjunctive 3.1.5 Verbal prefixes Adjectives 3.2.1 Inflectional forms; adjectival copula 3.2.2 Core forms 3.2.2.1 Conclusive and adnominal 3.2.2.2 Nominal 3.2.2.3 Exclamatory 3.2.2.4 The infinitives 3 .2. 2.4 .1 Infin itive-1 3 .2. 2.4 .2 Infinitive-2 3.2.3 Analytic forms 3.2.4 Ku versus shiku adjectives 3.2.4.1 Reduplication 3.2.4.2 Jiku adjectives Copula 3.3.1 Use of the simple inflected copula forms 3.3.2 Analytic forms Conjugation classes and morphophonology 3. 4.1 Regular verb classes 3.4.1.1 Quadrigrade verbs 3.4.1.2 Bigrade verbs 3.4.1.3 Formation of basic inflected forms 3.4.2 Irregular consonant base verbs 3.4.2.1 r-irregular 3 .4.2.1.1 Grammatical uses of ar- and other existential verbs 3 .4.2.1.2 Fused forms; secondary conjugations 3.4.2.2 n-irregular 3.4.3 Irregular vowel base verbs 3.4.3.1 Upper monograde 3.4.3.2 k-irregular and s-irregular 3.4.3.2.1 Grammatical uses of/w-and se-Kose3.4.3.3 3.4.4 Extended inflectional forms; combinatory stems 3.4.4.1 The infinitive as stem 3.4.4.2 The a- stem 3.4.4.3 The exclamatory as stem 3.4.4.4 The stative and the nominal 3.4.5 Consonant versus vowel base morphophonology

69 70 70 71 72 72 74 76 78 79 79 80 82 83 84 84 85 85 86 90 90 92

93 93 94 95 96 97 97 97 99 101 101 103 104 105 106 106 107 108 108 109 109 111 112 113 113

Contents

X

3.4.6

3.5

3.6 3.7

3.8

4

Loanwords 4.1 4.2

4.3

5

The kaJSU)IOkei system 3.4.6.1 Names and uses of the six kaJsuyokei 3.4.6.2 Traditional names for the verb classes 3.4.6.3 Discussion; basic paradigms Proto-Japanese and pre-Old Japanese morphology Verb classes; bigrade verbs 3.5.1 3.5.1.1 Diachronic classification of verbs 3.5.2 Pre-history of verb suffixes Verb extensions Particles 3.7.1 Case particles 3.7.1.1 Main Old Japanese case particles 3.7.1.1.1 Genitives 3.7.1.1.2 Differences between no and ga 3.7.1.2 Case marking of subject and object 3.7.1.3 Obsolete and peripheral case particles 3.7.1.4 Emerging case particles 3.7.2 Topic and focus particles 3.7.3 Restrictive particles 3.7.4 Conjunctional particles 3.7.5 Final particles 3.7.6 Interjectional particles 3.7.7 Complementizer 3.7.8 Etymology 3.7.8.1 Nominal sources 3.7.8.2 Verbal sources 3.7.8.2.1 Copula 3.7.8.2.2 Roots of other verbs 3.7.8.3 External etymology Pronouns 3.8.1 Short versus long forms 3.8.2 Personal pronouns 3.8.2.1 Other terms of address 3.8.3 Demonstratives 3.8.4 Basic pre-Old Japanese pronominal system 3.8.5 Proto-Japanese demonstratives

Ainu Continental loanwords: Korean, Chinese, Sanskrit 4.2.1 Korean 4.2.2 Chinese 4.2.3 Sanskrit Phonological adaptation

Eastern Old Japanese 5.1 5.2

Phonology Morphology

114 115 115 116 118 118 119 120 123 124 125 126 126 128 129 131 132 132 132 133 133 133 134 134 134 135 135 135 135 136 136 138 139 139 142 142

144 145 146 147 147 148 150

151 152 152

Contents

xi

Part II Early Middle Japanese 6

Writing and sources 6.1

6.2

7

Writing 6.1.1 Kanji-kana majiribun 6.1.2 Kana (hiragana, katakana, hentaigana) 6.1.2.1 Sei 'on and daku'on 6.1.2.2 Dakuten 6.1.2.3 Handakuten 6.1.2.4 Orthographic categories; the Iroha-uta, the Japanese 'alphabet' 6.1.2.5 New sounds Bound moras 6.1.2.5.1 Syllables with complex onsets 6.1.2.5.2 Syllable final/-tJ 6.1.2.5.3 Recent Modem Japanese loanwords 6.1.2.5.4 6.1.3 Orthographic norms 6.1.3.1 Undoing the etymological kana-spelling 6.1.4 Sound tables Sources 6.2.1 Prose and poetry 6.2.2 Annotated texts 6.2.3 Glossaries and dictionaries 6.2.4 Sanskrit studies

Phonology 7.1

7.2 7.3

Syllable structure 7.1.1 Bound moraic segments 7.1.2 Nasality 7.1.2.1 Morpheme internal position 7.1.2.2 Morpheme final position; postnasal neutralization 7.1.2.3 Word final position 7.1.2.4 Nasality harmony 7.1.3 Transcription ofmoraic segments 7.1.4 Sources of long syllables: the onbin sound changes Onbin as sound changes 7.1.4.1 7.1.4.2 Syllable reduction 7.1.4.3 Nasality 7.1.4.4 Major class; consonant or vowel? 7.1.4.5 Other sources of bound moraic phonemes The sound shape of Sino-Japanese vocabulary 7.2.1 Renjo Regular segmental sound changes 7.3 .1 Changes affecting OJ /p/ 7.3.1.1 Merger of intervocalic 1-p-/ with 1-w-1 7.3.1.2 Retention of 1-p-1 7.3.1.3 The Early Middle Japanese sound value of the reflexes of initial/p-I 7.3.1.4 Summary

157 157 157 158 162 163 165 165 169 169 170 170 170 171 175 177 178 179 181 182 183

184 185 187 188 188 189 190 190 191 191 195 196 197 199 199 199 200 201 201 202 203 204 205

Xll

Contents 7.3.2

7.4

8

Grammar 8.1

8.2

8.3 8.4

8.5 8.6 8.7

8.8

8.9

9

Loss of labial and palatal glides 7.3.2.1 Merger of IW-rui and otsu-rui syllables; loss of post consonantal/y/ before /e/ and lwl before /i, o/ 7.3.2.2 Loss of syllable initially/ before lei Loss of syllable initiallw/ before /o, i, e/ 7.3.2.3 1-i, -u/ versus /-.e, -.o/ 7.3.2.4 Phonemicization: /-.e, -.o/ or /-.ye, -.wo/ 7.3.2.5 Prosody; 'accent' 7.4.1 Prosodic classes 7 .4.1.1 Final falling pitch: classes 2.5, 3.5b, and 1.2 7.4.1.2 Initial rising pitch: class 1.3b; classes 2.3-5b? 7.4.2 Eleventh-century prosodic classes and later changes 7.4.3 Phonological interpretation 7.4.4 Complex forms 7.4.4.1 Noun + noun compounds 7.4.4.2 Particles 7.4.4.3 Verbs and adjectives

Verbs 8.1.1 Morphological categories 8.1.2 Conjugation classes and basic paradigms 8.1.3 Verbal nouns 8.1.4 Consonant base verbs Adjectives and copula 8.2.1 Secondary conjugations: ar- extended forms 8.2.2 Adjectival copula and negative 8.2.3 Copula Adjectival nouns Auxiliaries 8.4.1 Passives and causatives 8.4.2 Aspect; stative and perfective 8.4.2.1 Progressives Modality Derivatives Particles 8.7.1 Genitives 8.7.2 Conjunctional particles Tote, nado 8.7.3 Pronouns and demonstratives 8.8.1 Demonstratives and interrogatives 8.8.2 Short and long forms

Kakari-musubi 8.9.1 Uses of kakari-musubi 8.9.2 The individual kakari particles 8.9.3 The musubi predicate 8.9.4 Questions and exclamatives

The sinification of Japanese 9.1

Konbun-kundoku Kunten 9.1.1

205 206 206 206 207 208 210 212 215 216 217 219 222 222 223 224

227 227 227 227 229 230 232 233 233 234 235 236 236 238 240 240 242 242 243 244 245 245 246 247 247 251 251 255 256

258 258 259

Contents 9.1.2

9.2

9.3

Part III 10

Kanbun-kundoku and writing in Japanese 9.1.2.1 Hentai kanbun; 'kanbun' Orthographic overdifferentiation Kokuji Kuntengo The influence of kuntengo on the Japanese language

9.1.3 9.1.4 9.1.5 9.1.6 Ondoku 9.2.1 Chinese as a foreign (reading) language; Japano-Chinese 9.2.1.1 The phonetics and phonology of Japano-Chinese 9.2.1.2 Early Middle Chinese, Japano-Chinese and ongana 9.2.2 Sino-Japanese 9.2.2.1 Present-day Sino-Japanese 9.2.2.2 Syllable initial consonants Syllable final consonants 9.2.2.3 9.2.3 Sino-Japanese loanwords 9.2.3.1 Sino-Japanese loanwords in Old Japanese 9.2.3.2 Sino-Japanese loanwords in Early Middle Japanese 9.2.3.2.1 Numerals Sino-Japanese loanwords in Late Middle 9.2.3.3 Japanese 9.2.3.4 Sino-Japanese loanwords, Japano-Chinese, and Sino-Japanese Sino-Japanese words arising through on-reading of a 9.2.4 kun-writing The case of 1!f

263 266 268 268 268 270 274 274 277 277 278 280 282 282 284 284 286 289 290 290 291 292

Late Middle Japanese Sources 10.1 10.2

11

Xlll

Early Late Middle Japanese Late Late Middle Japanese 10.2.1 Shomono 10.2.2 Christian sources 10.2.2.1 Texts 10.2.2.2 Grammars 10.2.2.3 Dictionaries

Phonology 11.1

11.2 11.3

Nasality and medial voicing 11.1.1 Postnasal neutralization 11.1.1.1 'Vmu no xita nigoru' 11.1.2 Redistribution of phonemic nasality Loss of /w/ Fricativization of /p/; !pi> If/ 11.3.1 Retention of /p/ after /Q/ 11.3.2 Retention of /p/ after IN/ 11.3.3 Alternations arising from changes of /p/ 11.3.4 !pi in expressive vocabulary 11.3.5 Phonemic split: /p/ > If, pi

297 297 299 299 299 301 302 302

304 306 307 308 309 310 311 313 314 314 315 316

XIV

Contents Sino-Japanese syllable final 1-tJ 11.4.1 Sources of final 1-tJ 11.4.2 Renjo 11.5 Long vowels 11.5.1 Phonemic analysis 11.5.2 Sources of long vowels in Modem Japanese Sino-Japanese 11.6 Assibilation and palatalization 11.6.1 Palatalization 11.7 Final vowel deletion 11.8 Onglides 11.8.1 Diphthongal asymmetry 11.4

12

Grammar 12.1

12.2

12.3

12.4 12.5 12.6

Verb morphology 12 .1.1 Inflected verb forms 12.1.1.1 Gerund 12.1.2 Loss of exclamatory, conclusive and adnominal 12.1.3 Loss and change of auxiliaries 12.1.3.1 Morphological categories 12.1.3.2 From auxiliary to flective; inflection for tense 12.1.3.3 Non-finite forms 12 .1.4 Conjectural 12.1.4.1 Negative conjectural 12.1.5 Other auxiliaries 12.1.5.1 Negative 12.1.5.2 Passive and causative 12.1.5.3 New auxiliaries 12.1.6 Extensions Adjectives and copula 12.2.1 Adjectives 12.2.2 Copula 12.2.3 Summary of the development of regular and adjectival copula paradigms Basic paradigms 12 .3 .1 Verb classes 12.3.2 Onbin stems 12 .3 .3 Morphophonological rules Existential verbs; ar-, i-, or12.4.1 Loss of -ru Pronouns and demonstratives Syntactic changes 12.6.1 Loss of the distinction between conclusive and adnominal 12.6.1.1 Adnominal clauses 12.6.1.1.1 Complementizers 12.6.1.1.2 Tolwro-no 12.6.1.2 Loss of kakari-musubi 12.6.1.2.1 Namu andya 12.6.1.2.2 Ka and zo 12.6.1.2.3 Koso 12.6.1.3 Nominalizations; emergence ofnominalizing no

316 317 318 319 320 321 322 322 323 323 324

326 326 326 327 328 330 331 332 332 334 335 336 336 337 337 339 339 340 341 343 344 346 347 348 350 353 353 354 354 355 355 358 358 359 359 361 362

Contents 12.6.1.4

Summary of the developments of the functions of the Old Japanese/Early Middle Japanese adnominal 12.6.2 The genitive particles, subject marking and the emergence of a nominative case particle 12.6.2.1 Socio-linguistic differentiation of no and ga 12.7 Honorific language 12.7.1 Exaltation 12.7.1.1 Noun exaltation 12.7.1.2 Predicate exaltation 12.7.2 Politeness

365 366 367 369 369 370 371 372

Part IV Modern Japanese 13

Varieties of Modem Japanese 13.1

The Edo period: Linguistic diversity and common language 13.1.1 Common language 13.1.2 The language ofEdo 13.2 The Meiji period: Unification and standardization 13.2.1 Genbun'itchi 13.2.2 A national language; standardization and dialect eradication 13.3 Sources 13.3 .1 Material in alphabet writing 13.4 From Late Middle Japanese to standard contemporary Modem Japanese

14

Phonology 14.1 Merger of /d/ and /z/ before /i, u/ 14.2 Merger of I'.Y.JI and fool; h:-JI > fool 14.2.1 Nul diphthongs 14.3 Delabialization of Iff; Iff> lhl 14.4 Loss of phonetic onglides 14.4.1 Loss of palatalization before lei 14.5 Summary of main allophonic variation 14.6 Recent phonemic changes due to loanwords

15

Grammar 15.1

Verbs 15.1.1 Formation of the volitional 15.1.2 Verb classes 15.1.2.1 Levelling of vowel base verb conjugations; merger of monograde and bigrade verbs 15.1.2.2 Merger of n-irregular and quadrigrade verbs 15.1.3 Summary of verbal conjugation classes 15.2 Adjectives and copula 15.3 Other new Modem Japanese grammatical forms

16

Eastern dialect features of the standard language 16.1 16.2

Onbin forms Copula da

377 377 377 378 379 379 380 381 382 382

384 384 385 386 386 387 387 387 388

390 390 390 391 391 392 392 3 93 395

397 398 400

XVI

Contents 16.3 16.4

17

The negative auxiliary The imperative

The westernization of Japanese: Loanwords and other borrowings 17.1

Vocabulary layers and hybrid words 17 .1.1 Gairaigo 17.2 Pre-Meiji; from the end of Late Middle Japanese to the middle of the nineteenth century 17.3 From the beginning of Meiji to the end of WWII (1945) 17 .3.1 Loanwords from European languages 17 .3.2 Loan translations; Sino-Japanese coinages 17 .3.3 Influence from European languages in grammar and usage 17.4 Post-WWII

401 402

403 403 404 404 406 408 408 410 411

Appendix Summary of the main regular phonemic changes between Old Japanese and conservative Modern Japanese

413

References Index of main grammatical forms General index

416 425 429

Tables, maps and figures

TABLES

Table 1.1.

Important OJ sources

23

Table Table Table Table Table Table

Orthographically distinct syllables in early OJ Transcription systems for OJ Examples of transcribed OJ forms Phonemic syllables in OJ Phonetic realization of OJ obstruents Phonemic correspondences between OJ and pJ

27 32 33 33 35 42

Inflected forms for the eight OJ verb classes OJ auxiliaries attached to verbs from the eight verb classes Inflected forms of OJ auxiliaries OJ transitional paradigm and EMJ suppletive paradigm of the negative Reformation of the OJ paradigm of the negative OJ -n- perfective and negative forms of QD, LB and UB verbs OJ modal past and simple past auxiliaries OJ adjectival copula forms OJ copula forms Formation of OJ inflected verb forms OJ forms ordered according to formation on katsuyokei Katsuyokei paradigms for OJ verbs Katsuyokei paradigms for negative, simple past, adjectival copula and subjunctive Basic paradigm for OJ verbs OJ adjectival copula and possibly related grammatical forms Inflected forms of OJ verb extensions

54

2.1. 2.2. 2.3. 2.4. 2.5. 2.6.

Table 3.1. Table 3.2. Table 3.3. Table 3.4. Table 3.5. Table 3.6. Table Table Table Table Table Table Table

3.7. 3.8. 3.9. 3.10. 3.11. 3.12. 3.13.

Table 3.14. Table 3.15. Table 3.16.

60 61 70 71 73 74 81 94 100

llO l14 l17 l18 122 124 xvii

XVlll

Tables, maps and figures

Table 3.17. Table 3.18.

Pronominal fonns of OJ OJ demonstrative fonns

136 141

Table 6.1. Table 6.2.

159 161

Table 6.3. Table 6.4.

Kanji origins of kana Early kana shapes From Tsukishima (1977: 98) Phonemic transcription of the historical kana-spelling Important EMJ sources

Table Table Table Table Table

7.1. 7.2. 7.3. 7.4. 7.5.

EMJ sound inventory Examples of onbin Prosodic classes Phonological interpretation of the EMJ pitch patterns Prosodic shapes of EMJ basic inflected verb fonns

185 193 218 221 225

Table Table Table Table Table Table Table

8.1. 8.2. 8.3. 8.4. 8.5. 8.6. 8.7.

228 228 229 231 233 233

Table 8.8. Table 8.9. Table 8.10.

EMJ inflected verb fonns Basic paradigm of EMJ vowel base verbs Basic paradigm of EMJ CVi- and CVe- base verbs Basic paradigm of EMJ consonant base verbs EMJ adjectival copula fonns EMJ copula forms Basic paradigm of EMJ secondary conjugations of the copula, the adjectival copula and the negative auxiliary EMJ fonns of causative and passive auxiliaries Basic paradigm of the EMJ stative auxiliary -(i)tarEMJ demonstrative and interrogative forms

Table Table Table Table

Kanji readings EMC syllable initial consonants EMC syllable final consonants Native Japanese and SJ numerals

281 282 283 289

Table 10.1.

Important LMJ sources

298

Table 11.1. Table 11.2.

LMJ sound inventory Sources of SJ koo and yoo

305 321

Table Table Table Table

LMJ inflected verb fonns Merger of EMJ conclusive and adnominal Loss and changes of EMJ auxiliaries Changes of EMJ auxiliaries to LMJ ftectives

327 329 330 333

9.1. 9.2. 9.3. 9.4.

12.1. 12.2. 12.3. 12.4.

176 180

234 238 239 246

Tables, maps and figures Table 12.5. Table Table Table Table Table

Main fonns of the negative at the end of the LMJ period 12.6. Late LMJ adjectival copula fonns 12.7. Late LMJ copula fonns 12.8. Development of the copula and adjectival copula paradigms 12.9. LMJ basic verb paradigms 12.10. Summary of the development of the uses of ga and no

XlX

336 340 342 344 345 368

Table 14.1.

Free moras in conservative cNJ

385

Table 15.1. Table 15.2.

Standard cNJ inflected verb forms Changes among verbal conjugation classes from OJ to NJ cNJ adjectival copula forms cNJ copula fonns

391

Table 15.3. Table 15.4. Table 17.1. Table 17.2.

393 394 395

Loanwords in use during the Taisho period From Shibatani (1990: 149, table 7.5) Loanwords used in magazines From Shibatani (1990: 148, table 7.4)

408

Distribution of prosodic systems From Shibatani (1990: 211) Major east-west isoglosses From Shibatani (1990: 197) Distribution of words for 'pumpkin' From Sato (2002: 169)

213

411

MAPS

Map 7.1 Map 16.1 Map 17.1

399 407

FIGURE

Figure 9.1

Example of Kunten system

260

Acknowledgements

I would like to acknowledge the support, both funding for research travel and research assistance and support in the fonn of sabbatical leave, that I have received in the course of writing this book from the Faculty of Oriental Studies and Hertford College in the University of Oxford. I also want to express my gratitude to the University's Sasakawa Fund and John Fell Fund for their financial support. Without the generous support of these institutions, I would not have been able to complete this project. A number of present and fonner Oxford students provided research and editorial assistance and valuable comments on parts of the book: Madeleine Brook, Marion Cossin, Laurence Mann, Dan Millichip, Nina Mirnig, Jenny Moore, Cecily Nowell-Smith, Thu Phuong Nguyen, Muneto Ozaki, James Stone-Lunde, James Webb, and Asa Yoneda. On numerous occasions Greg Jennings and his staff at the IT department in Hertford solved hardware and software problems for me. Anton Antonov, Stephen Wright Horn, Kerri L. Russell, Matt Shibatani, Antoni U<;erler, Uwano Zend6, and an anonymous reader offered important comments and suggestions on earlier drafts of the book. Among the many colleagues from whom I have learned much of what is reflected in this book, I must also in particular mention Kinsui Satoshi, Alexander Vovin, and Janick Wrona. Janick also provided invaluable comments and suggestions for improvement on the final draft. Above all, John Whitman has been instrumental in shaping my views on general and specific problems in the history of the Japanese language. John read and commented extensively and incisively on several draft versions and much of the value this book has in its present fonn is due to his input. Naturally, all insufficiencies that remain, despite the best efforts of those mentioned above and many others as well, are entirely of my own making. I would also like to thank Andrew Winnard, Sarah Green, Joanna Garbutt and Rosina DiMarzo of Cambridge University Press for their guidance and help in producing the book. Anna Oxbury was an amazing copy-editor. I am grateful to my father-in-law for the calligraphy of the Iroha-uta on the cover of the book. Finally, for emotional and practical support through the long process of putting together and completing this book, I have my children and my wife to thank. XX

Abbreviations

SYMBOLS

/ .. ./

II... I/ => >

syllable boundary enclose phonemes or phonemic representation enclose underlying representation generative rules ('becomes in the course of derivation') diachronic correspondence ('becomes through time')

GRAMMATICAL TERMS

ABL ACC ACOP ADJ ADN ALL AUX CAUS COM COMP CONC COND CONCL CONJ CONT COP DAT DESID EMPH ETOP EVID EXCL FOC

ablative accusative adjectival copula adjective adnominal allative auxiliary causative comitative com plem entizer concessive conditional conclusive conjectural continuative copula dative desiderative emphatic emphatic topic evidential exclamato:ty focus xxi

XXll

GEN GER H

HON HUM IMP

INF INT

intr. k-irr L

LB LM MPST MVR NCONJ NEC NEG

n-1rr NMLZ NMNL NOM NONPST OPT PASS PCONJ PERF POL PRES PROV PST PSTCONJ PURP Q

QD RESP

r-1rr S-liT

SPST STAT SUBJ

Abbreviations

genitive gerund high (tone or pitch) honorific humble imperative infinitive intentional intransitive k-irregular (ka-hen) low (tone or pitch) lower bigrade (shimo nidan) lower monograde (shimo ichidan) modal past mid vowel raising negative conjectural necessitive negative n-irregular (na-hen) nominalizer nominal nominative nonpast optative paSSIVe present conjectural perfective polite presumptive provisional past past conjectural purpOSIVe interrogative quadrigrade (yodan) respect r-irregular (ra-hen) s-irregular (sa-hen) simple past stative subjunctive

Abbreviations TOP

tr. UB UM VOL

xxiii

topic transitive upper bigrade (kami nidan) upper monograde (kami ichidan) volitional

LANGUAGES

cNJ EMC EMJ EOJ J-Ch LMC LMJ MC MJ :MK

NJ

oc OJ pJ pK SJ Skt

contemporary Modem Japanese Early Middle Chinese Early Middle Japanese Eastern Old Japanese Japano-Chinese Late Middle Chinese Late Middle Japanese Middle Chinese Middle Japanese Middle Korean Modem ('new') Japanese Old Chinese Old Japanese proto-Japanese proto-Korean Sino-Japanese Sanskrit

TEXTS

Ars gr. Arte Arte breve Bussoku EN Esopo Feiqe Genji Ise KK Kokin MYS NSK

Ars grammaticae iaponicae linguae Arte da lingoa de Iapam (page references are to Doi 1955) Arte breve da lingoa Iapoa Bussokuseki-ka (Engishiki) Norito Esopono fabulas Feiqe monogatari Genji monogatari Ise monogatari Kojiki kayo (songs/poems in the Kojiki) Kokinwakashil Man'yoshil Nihon shoki kayo (songs/poems in the Nihon shoki)

XXlV

Abbreviations

Ochikubo SM Take tori Tosa Vocabulario

Ochikubo monogatari (Shoku nihongi) Senmyo Taketori monogatari Tosa nikki Vocabulario da lingoa de Iapam

Introduction

This book describes the development of the Japanese language from its written beginning until the present day as it is reflected in the written sources; that is to say, its internal history. This is accomplished by first giving an overall description of the oldest attested stage of the language, Old Japanese, and then tracing changes since then, as they are reflected in the written sources and in the present-day language. The possible cognation of Japanese to other languages, its external pre-history, is not considered. Nor is its internal pre-history discussed to any significant extent, except where relevant to understanding its attested history. And nothing is said about dialects, except where they are prominently reflected in the written sources and where they have contributed to the formation of the modern standard language. The periodization of Japanese shown in (1) is adopted, which overlaps with the main political periods. (1)

Linguistic periods Old Japanese (OJ) 700-800

Political periods Nara, 712-794

Early Middle Japanese (EMJ)

800-1200

Heian, 794-1185

Late Middle Japanese (LMJ)

1200-1600

Kamakura, 1185-1333 Muromachi, 1333-1573

Modern Japanese (NJ)

1600-

Edo, 1603-1868 Meiji, 1868-1912 Taish6, 1912-1926 Sh6wa, 1926-1989 Heisei, 1989-

Where necessary, early is distinguished from late within both Early Middle Japanese (early: 794-1086; late: the Insei period, 1086-1185) and Late Middle Japanese (early: the Kamakura period; late: the Muromachi period). Modern Japanese is abbreviated as 'NJ' (for 'new Japanese') to avoid

2

Introduction

confusion with :Middle Japanese (MJ). By 'cNJ' is meant contemporary NJ, the Japanese of the twentieth century. 'Classical Japanese' is the fossilized, relatively fixed written norm which arose largely out of the language of the twelfth century and which thereafter remained the dominant form of writing in Japanese until the beginning of the twentieth century. This book is not to any great extent concerned with the fossilized Classical Japanese written norm. Extensive attestation of Japanese goes back to the eighth century. Old Japanese is mainly the language of the Nara period, although it also comprises earlier texts which are included in sources compiled or completed in the eighth century, but their age is difficult to assess. It is sometimes said that Japanese has not changed greatly since Old Japanese. However, it is possible to identify two large sets of internal change, largely coinciding with EMJ (phonological change) and LMJ (grammatical change), respectively, as well as two waves of contact induced change which took place during OJ!EMJ (sinification) and late NJ (westernization), which together transformed the language from its Old Japanese to its modern form. In the transition between OJ and EMJ and within EMJ, the language underwent significant phonological changes, both in syllable structure and in segmental phonology. At the end of the EMJ period, the phonological structure of Japanese was largely as it is today and phonological changes since then have by comparison been minor. Most major phonological changes were complete by the end of the eleventh century and some scholars do not include the last century of the Heian period, the so-called Insei period (1 086-1185), as a part ofEMJ, but instead as part ofLMJ. It was also during the EMJ period that the contact with Chinese manifested its influence in the texts, although we suspect that this influence was present already in OJ. During the LMJ period by contrast, major grammatical changes took place, which affected both morphology and syntax. Some of these changes are initiated or anticipated in the twelfth-century materials, but they are mainly reflected in the written sources of the LMJ period. The dating of the end of LMJ is difficult to determine on linguistic grounds. The main issue is whether to include the language stage reflected in the Christian materials from the end of the sixteenth century and early years of the seventeenth century in LMJ or in NJ. The sweeping grammatical changes which took place during LMJ were complete by then and in many respects the language of those sources is similar to NJ. However, it is also very similar to the language in an earlier set of sources, the so-called shomono which date from the middle of the fifteenth to the middle of the sixteenth century. The Christian sources, alongside the shomono, are accordingly included in the LMJ period, and we will consider them to constitute the end of LMJ. NJ is thus the language from then on. Once the phonological changes of EMJ and the grammatical changes of LMJ were complete, Japanese did not

Introduction

3

change very much structurally. However, NJ was affected by significant external changes brought about by extensive contact with European languages in the course of the modernization of Japan and Japanese in the late nineteenth and early twentieth centuries.

Attestation of forms OJ, EMJ and LMJ are dead languages and the sources are closed and limited text corpora. Naturally, many specific forms of individual words are not attested in those texts. The facts and state of attestation are of paramount importance when investigating the system of a language, but of less interest when explaining or exemplifying the system. Recall Winston Churchill's famous experience, recounted in the first chapter ofhisMy Early Life, ofbeing set the task upon arrival at his boarding school at the age of seven of learning the singular of the noun of the first declension in Latin. He managed to memorize the paradigm and reproduce it to the satisfaction of his teacher, but he did not understand what it meant and asked about it: 'But,' I repeated, 'what does it mean?' 'Mensa means a table,' he answered. 'Then why does mensa also mean 0 table,' I enquired, 'and what does 0 table mean?' 'Mensa, 0 table, is the vocative case,' he replied. 'But why 0 table?' I persisted in genuine curiosity. '0 table,- you would use that in addressing a table, in invoking a table.' And then seeing that he was not carrying me with him, 'You would use it in speaking to a table.' 'But I never do,' I blurted out in honest amazement. 'If you are impertinent, you will be punished, and punished, let me tell you, very severely,' was his conclusive rejoinder.

An inflectional paradigm is a pattern of relations~ the combination of mutually exclusive inflectional endings with stems of words~ or, put differently, the morphologically possible forms of a word. It is possible that no one ever produced the vocative of mensa in actual speech or writing in classical Latin. But if anyone wished to address a table, the vocative case was available for that purpose and even if it was never spoken, the vocative form of that noun existed in Latin as a systemic possibility. In OJ, the imperative of the verb kog- 'row' is not attested. There is no systemic reason that kogye 'row!' should not exist, much less so than for the vocative of Latin mensa. The nonoccurrence of kogye may be regarded as an accident of attestation. Given the existence of OJ kog-, which is amply attested in various forms, we know that its imperative would be kogye. On the other hand there are restrictions on the use of the imperative of certain types of verbs, e.g. those which signify spontaneous, nonvolitional action. It is debatable whether this is a morphological restriction (that form does not exist) or a syntactic/pragmatic restriction

4

Introduction

(that form cannot be used, it does not make sense for that fonn to be used), but it is a grammatical restriction in the language, not an accident of attestation. Throughout the book morphological paradigms, particularly verbs, are exemplified. Some of the fonns of some of the verbs are not attested, but this is generally not remarked, except for small, irregular verb classes or grammatical auxiliaries where the situation of attestation may be of significance. Conventions

Throughout the book a phonemic transcription of cited forms is generally employed, appropriate to the period of citation. For example, the word for 'front', which is NJ mae, will be written as shown in (2), reflecting its phonemic shape at different stages of the language (exemplified by texts from those periods): (2)

OJ Man 'yoshu mapye

early EMJ Tosa nikki mape

mid EMJ Genji monogatari mawe

late EMJ onwards Sarashina nikki mae

Japanese editions of pre-modem texts, by contrast, employ a historical spelling and will spell 'front' as~"" ('mahe') regardless of the period from which the text dates. From late EMJ until the second half of NJ this word was, like all words of the shape /Ve/, pronounced with a palatal onglide before the /e/, [maie], but that will not be noted in the phonemic transcription. Nor will other allophonic features of pronunciation, such as the prenasalization of /b, d, g, zl (which from OJ into the LMJ period were pronounced ["b, nd, ug, nz]). However, when transcribing forms with a moraic consonant (which in EMJ and early LMJ was phonemically underspecified for nasality in morpheme internal position), we will use a semi-allophonic transcription and write punde 'brush', not puCdE, although strictly speaking the phonemic shape was /puCde/; see further 7.1.3 about this. When citing words or passages from OJ texts italics are used for phonographically written text portions and nonnal type for logographically written text. For example, when citing the word suru, the adnominal fonn of se- 'to do', it is transcribed in accordance with the writing in the source as exemplified in (3), with different writings of suru cited from different poems in the Man 'yoshu. In (3a), ~J{ and rfrt are used phonographically for the syllables /su/ and /ru!. In (3b), ~ is used logographically to write part of a form of a verb meaning 'to do' and :ofE is used phonographically for the syllable /ru!. Finally, in (3c), ~ is used logographically for a fonn

Introduction

5

of a verb meaning 'to do' which reading tradition interprets as the adnominal form. (3) a. b. C.

MfSpoem no. 17:3932 19:4198 2:164

Writing ~J{¥frE

~mE ~

Transcription suru surn suru

Words from cNJ will be noted in the modified Hepburn transcription, for example fuji '(Mount) Fuji' or choshoku 'breakfast'. Japanese script will generally not be used, except when discussing script and writing. However, where relevant the Chinese characters (kanji ~*) used to write Chinese or SinoJapanese words are given. Japanese proper names will usually be given in their usual cNJ form, e.g. Hojoki (not the shape this word would have had at the beginning of the thirteenth century: Pafidyafi-ki) and personal names are given in the traditional Japanese order (surname, given name), e.g. Hattori Shiro. Modern Chinese words will be transcribed in pinyin. Early Middle Chinese reconstructions follow Pulleyblank 1991. When citing verbs, the basic stem is used, rather than the OJ/EMJ/ early LMJ conclusive and the late LMJ!NJ nonpast form, which are the forms usually used as citation forms, i.e. the forms used to cite or talk about a verb, for example in dictionaries. 1 Citing the basic stem in the majority of cases unambiguously identifies the conjugation class of a verb, especially when comparing quadrigrade and bigrade verbs, e.g. OJ ok- 'put' (quadrigrade), ake- 'dawn' (lower bigrade), okwi- 'arise' (upper bigrade), whereas that is not the case with the OJ/EMJ/early LMJ conclusive: oku, aku, oku. Almost all verb suffixes attach directly to the basic stem of regular vowel base verbs, but for some irregular vowel base verbs and for all consonant base verbs, most verb suffixes attach to one of several derived stems (see further 3.4.4 and 8.1.4). When citing verbal suffixes a morphophonemic notation is used which shows which, if any, derived stem the suffix attaches to, by using bracketed prefixes: (a), (i), (e), (I). This is exemplified in (4), using the verb sak- 'to come into bloom'. The a- stem corresponds to the mizenkei of Japanese school grammar, the infinitive to the ren 'yokei, and the exclamatory to the izenkei (see 3.4.6). 1

Notable exceptions, which list verbs under their infinitive, are Ohno's dictionary of pre-modern Japanese (1990) or dictionaries published by the Jesuit missionaries in the early seventeenth century, e.g. Rodrigues (1603-4).

6

Introduction

(4)

Prefix (a) (i) (e) (I)

Stem a- stem Infinitive Exclamatory onbin stem

Verb sakasakisakesai-

Suffix -(a)n- (Negative) -(i)n- (Perfective) -(e)do (Concessive) -(I)ta (Past)

Verb + suffix saka-nsaki-nsake-do sai-tar-

This notation does not mean that the prefixes are part of the (synchronic) phonemic shape of the suffixes. The prefixes only show which stem a suffix selects. For example, for both the negative, -(a)n-, and the perfective, -(i)n-, the basic stem shape of the suffix is /-n-/, but they select different stems, as shown in (4). In examples, verb forms will be segmented as in (5), i.e. noting (by '-') morpheme boundaries between verb stems and auxiliaries, but not between ftectives and the immediately preceding verb or auxiliary stem. In glosses, the inflected form will be noted as part of the gloss for a verb or auxiliary, separated by '. ', showing that saku is the conclusive form of the verb sak- 'bloom', sake do is the concessive form of that verb, and -kyeri is the conclusive form of the modal past auxiliary -(i)kyer-. This notation shifts the focus away from individual morphemes to the actual inflected word forms. See sections 3.1 and 3.4 for the analysis underlying this notation. (5) a.

saku bloom.CONCL 'it blooms'

b.

sakedo bloom.CONC 'although it blooms'

c.

maywopi-ki-ni-kyeri fray-come-PERF-MPST.CONCL 'it had become frayed'

Where possible, examples from Japanese texts are cited from Iwanami's critical edition Nihon koten bungaku taikei (Iwanami 1957-69). Examples from Nara periodSenmyo (Imperial edicts, see 1.2.3.2) are cited from Kitagawa (1982) and texts originally published in print, such as the Clrristian materials from the end of the sixteenth and beginning of the seventeenth centuries (see 10.2.2), are cited from the originals (or photographic reproductions). References to Rodrigues'sArte da lingoa de Iapam (1604-8) are to Doi's (1955) translation into Japanese, which is more easily accessible for most readers and which gives page references to the original.

Introduction

7

As few references as possible are given in the body of the text, but a short list of references is provided at the end of most chapters; where possible references are given to scholarship in English. Overall, factual information or generally accepted descriptions which are available in common handbooks, overviews and dictionaries, or by looking at the texts, are not referenced. Handbooks frequently consulted include: The Japanese language through time (Martin 1987), Jidai-betsu kokugo daijiten: Jodaihen (Omodaka et al. 1967), Kokugogaku daijiten (Kokugogakkai 1980), Kokugogaku jiten (Kokugogakkai 1955), Nihon kokugo daijiten (Shogakukan 2000-2), Kokugogaku kenkyi{jiten (Sat6 1977), Nihon bunpo daijiten (Matsumura 1971 ), Nihongo bunpo daijiten (Yamaguchi andAkimoto 2001), Nihongogaku kenkyujiten (Hida et al. 2007), Nihongo hyakka daijiten (Kindaichi et aL 1988). Of these, the Nihongogaku kenkyujiten has an exhaustive listing and description of available textual sources (pp. 629-1129).

Part I

Old Japanese

1

Early writing in Japan and Old Japanese sources

1.1

Writing

1.1.1

Introduction of writing in Japan

The Japanese were exposed to written matter as early as the late Yayoi period (c. ?1000 BC- 300 AD). Thus, inscribed Chinese coins have been unearthed in excavations of sites dating from the first century AD. There is no evidence of any awareness of the function of writing at that early stage, however, and it is likely that the characters which appear on mirrors and other artefacts produced in Japan through the third and fourth centuries were also simple ornaments, in imitation of those found on articles from the continent. To all appearances, writing as such, in the form of Chinese Classics, was introduced into Japan early in the fifth century as part of the great cultural influx from Paekche. The Kojiki and the Nihon shoki recount this event as the advent of the scribes Wani and Akichi in the years Ojin 15 and 16 (thought to be early in the fifth century, possibly 404-5~ the traditional dating puts this at as early as 284-5, two 60-year cycles earlier). For some time, writing remained in the hands of hereditary professional scribes ifubito) who were of continental heritage. Through the sixth and seventh centuries Sinitic culture, including Chinese Buddhism, flowed into Japan through Paekche. In the course of this, written Chinese assumed enormous importance in matters of state, philosophy and religion. Any serious engagement with such matters required knowledge of written Chinese and for some time writing was equivalent with writing in Chinese. Also composition of Chinese poetry became highly regarded and remained so long into the medieval period. Thus, the oldest surviving poetry anthology in Japan is the Kaifiiso ('tli!J!U~t c. 751) a compilation of Chinese poetry written in Japan. Reading and interpretation of Chinese canonical texts came to assume great importance, both within Buddhism and in government administration. Chinese texts were read in two ways: either (a) reading them out in a form of Chinese (ondoku ifW'E 'sound reading')~ or (b) translating or rendering the texts into Japanese (kundoku ~}iiWt 'gloss reading'). These two practices have exerted great influence both on the Japanese language itself and on the way it is written. This will be discussed in detail in 9.1. 11

12

1 Early writing in Japan and Old Japanese sources

1.1.2

WritinginJapanese

The earliest attestation of writing in Japanese dates from the fifth century, but it is not until the middle of the seventh century that writing in Japanese became widespread. Throughout the OJ period Japanese was written entirely in kanji which were used logographically or phonographically.

1.1.2.1 Logographic versus phonographic writing Writing is a representation of language: elements of writing represent elements of language. Depending upon the nature of the linguistic elements that elements of writing stand for, there are in principle two types of writing. First, writing which represents those elements of language which carry meaning: words or morphemes. This is logographic writing. Second, writing which represents those elements of language which distinguish among elements carrying meaning: phonemes or phonological units of greater or smaller extent. This is phonographic writing. Below, these two types of writing are illustrated with examples from NJ. (la) shows logographic writing, with ~~ standing for the word which has the sound shape /tokil and the meaning 'time'; ~~ does not stand primarily for the meaning or the sound shape, but for the word, the linguistic sign, which comprises both. (lb) exemplifies phonographic writing, with ~ standing for the syllable /to/; ~ can thus be used to write any occurrence of /to/ regardless of the word of whose sound shape /to/ forms part. (1)

a.

~~

Iltokil I I 'time'l

b.

[I] ~

Actual orthographic systems and practices rarely, if ever, limit themselves to one of these types of writing. For example, Chinese, which is the stock example oflogographic writing, has a strong phonographic element, DeFrancis (1984) arguing that this is more prominent than the logographic element. Conversely, most writing systems have a logographic element. This includes alphabet writing as used to write English; for example, red and read (past tense of the verb 'to read') are written differently although they are homophonous. Also spaces between words, capitalization of some words, and punctuation all contribute an element of logography to alphabet writing.

1.1.2.2 Adaptation ofChinese script It is not known specifically when or how the Chinese script began to be used to write Japanese. Nor is it clear by what stages this took place. It is clear, though, that it makes little sense to consider this from a purely Japanese per-

1.1 Writing

13

spective. This development should be seen within the wider context of the adaptation within the Sinitic cultural sphere of Chinese script to write local languages, which took place earlier on the Korean peninsula than in Japan. There is ample evidence, in the form of orthographic 'Koreanisms' in the early inscriptions in Japan, that the writing practices employed in Japan were modelled on continental examples; later texts with Koreanisms include the epitaph from the grave of 0 no Yasumaro, the compiler of the Kojiki. Scholars from the Korean peninsula, particularly Paekche, played an essential role in the development of writing in Japanese by introducing, practising (as scribes) and teaching methods of writing already in use on the Korean peninsula. It is important to appreciate, however, that the presence of identifiable 'Koreanisms' in a text simply means that it incorporated writing practices developed on the continent; it says nothing about where or by whom the text was written. Chinese script may be adapted to write other languages either logographically or phonographically. Pre-alphabetic writing in Korean comprises the following three main types: (a) pure logographic writing, with kanji used for lexical words, but with no indication of grammatical particles or morphology; (b) logographic writing, with conventionalized logographic and phonographic writing of some grammatical elements (Korean idu 'clerk readings'); (c) logographically written lexical items supplemented by phonographically written grammatical elements (hyangch 'af). These types are all found in the OJ text corpus. It is likely that the correspondences in specific types of writing on the continent and in Japan reflect a common continental source rather than parallel developments and there is therefore little sense in trying to reconstruct an independent course of evolution of adapting Chinese script to write Japanese. However, writing extensive text passages entirely or mostly phonographically, reflected in the widespread use of man 'yogana (1.1.2.5), is a practice not attested in Korean sources which therefore seems to be an independent development which took place in Japan.

1.1.2.3 Logographic writing ofJapanese The principle for logographic adaptation of Chinese script is that you represent a word with a character which stands for a Chinese word which is semantically or referentially similar to the word you wish to write. In a sense, you translate the word you want to write into Chinese and use the character which stands for the resulting Chinese word. Simplistically put, you take the meaning associated with a character with no regard to its sound value. For example, -:*:, which in modem Chinese stands for the word quiin 'dog', could be, and today is, used to write the Japanese word inu 'dog'; or it could be used to write the English word dog. In an extreme type of logographic writing of Japanese, known as hentai kanbun 'deviant Chinese text', stretches of text longer than

14

1 Early writing in Japan and Old Japanese sources

single words are represented by strings of kanji in an order different from the word order in Japanese, making the texts appear more like Chinese than Japanese, but they are in fact complex representations of Japanese (see further 9.1.2.1).

1.1.2.4 Phonographic writing ofJapanese Phonographic writing makes possible a precise and unambiguous recording of linguistic forms, something which is not possible to the same extent in logographic writing. The principle for phonographic use of kanji is that a sound sequence- usually a syllable- is represented by a character otherwise used logographically to write a word whose sound shape is similar to the syllable you want to write. Simplistically, you take the sound value of a kanji with no regard to its meaning. For example, iJJF which stands for the word mi 'full, fill, more' in Chinese, could be used to write the syllable /mi/ in Chinese regardless of the word in which it occurs. Or it could be borrowed to write the sound sequence /mi:/ in English, or /mil in Japanese. Thus for example, in both Chinese and Japanese the loan word meaning '(Catholic) mass' borrowed in the late sixteenth century from Portuguese missa is written iJJF~, standing for Chinese misa, Japanese misa. In China itself there is a long tradition for transcribing foreign names and words whose sound shape is thought to be important in this way. That was for example the case with the religious terminology of Buddhism when it was introduced into China, 1 or with the recording of foreign words in Chinese histories. 1.1.2.5 Man'yogana Kanji used to write Japanese phonographically are termed 'man 'yogana' after their extensive use in the poetry anthology Man'yoshu (1.2.3.1). Sometimes the term 'man 'yogana' is mistakenly used to refer to the entire orthography of the Man 'yoshu, but this is wrong on two counts: First, 'man 'yogana' refers exclusively to phonographic writing and not to the logographically written parts of the Man 'yoshii. Second, man 'yogana are used also in other texts than the Man 'yoshii and the use of man 'yogana continued long into the Middle Japanese period, also after the emergence of the simplified kana scripts (cf. 6.1.2). It is customary to distinguish between two basic types of man 'yogana, depending on the provenance of their sound values: (a) ongana, which were used as phonograms on the basis of their (Japano-)Chinese readings; and (b) kungana, which were used on the basis of their kun-readings. This is illustrated 1

Buddhist texts were translated into Chinese, but important concepts and tenninology in them were rendered phonographically because their sound shape was thought to be related to their function, or because concepts such as nirvana were so alien as to make transliteration the only sensible option.

1.1 Writing

15

below with /'1( which was used both as an ongana and as a kungana. Logographically, this character wrote words meaning 'tree' and the like. It was used as a kungana for the syllables /ko/ and /kwi/ because the OJ word meaning 'tree' had the variants ko- - kwi. It was used as an ongana for the syllable !mol because the phonetic manifestation of the Chinese word meaning 'tree' was similar to the phonetic manifestation of the OJ syllable /mol. (2)

Logogram

*

Ongana

I,: ,1

[m;)wk]

0

'tree'

(3)

Logogram

Kung ana

*

l:vll:

/kwi- ko-/ 'tree'

Below are given some more examples of man 'yogana: (4)

Kanji

Ongana

Early Middle Chinese

!t

/pi/

HE

/pwi/ /myel /myel

*pjNbjih; pji'; bji *buj *mji;)' *mai:jh, msl *msj' *ko' *xi:::\' *tsi:' /tsi' *m;)wk *p;)i:t/ps:t *dsn

II$ ~

.?13

*if ~f ~

*83 )\.

/me/ /kwo/ /ko/ /sil /mol /pal /de/

Kungana

OJ word

lkwo/ lko, kwi/ /ya/ Ita/

kwo 'child' ko- ~ kwi 'tree' ya 'eight' ta 'paddyfield'

In a very general sense, kanji were used as ongana on the basis of a perceived phonetic similarity between the OJ syllables they represented and their

16

1 Early writing in Japan and Old Japanese sources

Chinese pronunciation. The adapted fonn of Chinese pronunciation used in Japan will be referred to as 'Japano-Chinese'; we will discuss this tenn and its implications in 9.2.1. There are two points to be made in that regard: (a) The phonetics of J-Ch is not known. Early Middle Chinese (c. 601; see 9.2.1.2) is available in a number of reconstructions (here we follow Pulleyblank's (1991) reconstruction), but the relation between EMC and J-Ch is not direct (see 9.2.1.2). (b) In addition to phonetic and phonological considerations, it is likely that scribal tradition brought along from the continent by scribes, as well as any number of other 'extra-phonological' factors, played an important part in the choice of kanji to act as ongana. Consider for example the sound values (including the NJ on-readings) of% (EMC: *mail/msl; Go-on: me; Kan-on: bm), * (*msj'; mai; bei), and ~~~ (not in Pulleyblank 1991, but reconstructed by Miyake (p.c.) as EMC *myi;;)q, which corresponds to Pulleyblank *mji;;)'; mi; bi). It is clear that% has sound values far more similar to* than to ll~, yet% is used, like ll~, as an ongana for the syllable /myel, as opposed to * which is used to write /me/.

1.1.2.6 Senmy6-gaki The majority of OJ texts are written in a mixture of phonographic and logographic writing. Whereas some mixed texts have no clear functional differentiation between phonograms and logograms, others mainly use phonograms to write grammatical elements and logograms to write lexical words. A distinctive way of writing which has become known as senmyo-gaki 'edict-writing' (after its use in the Senmyo, see 1.2.3.2), is a mixture of logographic and phonographic writing in which some grammatical elements were written phonographically in smaller size characters than the rest of the text. Senmyo-gaki is a refined writing system, showing a sophisticated grammatical understanding of the language. It may well have been inspired by the Korean idu (see 1.1.2.2). Senmyo-gaki is similar to the mixed writing of modem Japanese, in the sense that both exhibit a high degree of orthographic distinction between lexical words and grammatical elements. Senmyo-gaki is usually associated with Senmyo and Norito which are the main texts written in this way. It is, however, not exclusive to them, but is found in other texts as well. Both wooden tablets (mokkan) (see 1.2.2) and archival records from the mid eighth century show that the practice of distinguishing in size between characters used phonographically for grammatical markers and others was well established by that time, but it is not clear how far back this type of writing dates, and thus if all Senmyo and Norito were originally recorded in this way. Thus mokkan from the second half of the seventh century include texts in writing which otherwise is very similar to senmyo-gaki, but without a distinction in size.

1.1 Writing

1.1.3

17

Problems of decipherment

Writing is a representation of language and reading consists in reconstructing a linguistic specimen, a text, from its written representation, i.e. decoding the text. It thus requires knowledge of the 'code of transmutation' employed when a text was written down, i.e. encoded in writing. Although reading is commonplace and we tend to think little of it, it is in principle a complicated process. This is clear when we approach texts representing an older language stage or an unknown language with the purpose of establishing what the texts might mean or what the language might be like. This is very different from mapping our knowledge of a language onto a written representation of a text in order to recover it. Individual OJ texts exhibit some degree of consistency in their orthography, but viewed as a whole OJ writing is enormously complicated. This holds in particular for the writing in the main source of the OJ language, the Man 'yoshii, which is not a single text, but a compilation of texts. The complexity of OJ writing means that deciphering the OJ text corpus is no simple matter and there are in fact still many obscure points despite a long philological effort.

1.1.3.1 Polyvalence and equivalence The main regular problems of decipherment are posed by the polyvalence and equivalence of kanji. (a) Kanji used to write OJ were polyvalent in several respects. First of all, they could be used as logograms or as phonograms. As shown in the examples above, kanji used as phonograms could be used as ongana or as kungana and some were used as kungana for different syllables (this mostly confined to variant forms of one word). When used as logograms, kanji could be used for different OJ words. For example, :;t: (EMC *khi:;)h 'go away, depart') could stand for the words sar- 'leave' or yuk- 'go'. (b) On the other hand, some kanji were equivalent in the sense that one linguistic unit could be represented by several different kanji. For example, the word yuk' go' could be represented logographically with, amongst others, 1T (*yai:jl)/ys:jiJ 'walk, go'), lUI (*~ial 'pass away, die'), :t: (*khi;)h 'go away, depart'), or f.± (*wuaiJ' 'go'). The origin of such usage is the translational inequivalence between Chinese and Japanese: several OJ words could be used to translate :ii:, and vice versa several Chinese words could be translated by OJ yuk-. Finally, each syllable could be written phonographically by a number of different kanji; for example, in the Man 'yoshii close to twenty different kanji are used as man 'yogana for the syllable lka/. The problems of decipherment posed by polyvalence and equivalence cannot be solved on the basis of the texts alone. However, the main texts were read and copied through time and annotated in order to facilitate reading when in later periods the language and the orthography became obsolete. There is

18

1 Early writing in Japan and Old Japanese sources

thus a tradition concerning these texts which fonned the point of departure for a long and thorough philological effort of deciphennent; and today the phonographically written portions of the OJ texts are deciphered to general satisfaction on most points of significance.

1.1.4

Reading tradition

Reading tradition gives voice to the logographically written portions of the OJ texts. Undoubtedly the traditional readings are in many cases correct, in the sense that they reflect the words and word forms which were originally intended in a text. Strictly speaking, however, the readings constitute hypotheses about the texts; often probable hypotheses, well founded on sound knowledge of the language and on tradition, but all the same hypotheses which mostly remain unverifiable. Generally, we cannot know which words are represented in an older logographically written text. On this background it is obvious, but still enonnously important to make clear, that no argument about the phonology or morphology of OJ can be based on logographically written text portions and their traditional readings. They can provide valuable clues in matters of syntax, but also in this area caution must be exercised.

1.1.5

Rebus writing

A special and intriguing problem is posed by obscure rebus writings, employed in particular in the Man 'yoshii. Two well-known examples will suffice here. The first is found inMYS9.1787 where a fonn of the word ide- 'emerge, come out' is written by the five characters LlL..I::.tii~Lll which stand for words meaning 'mountain top again exist mountain', in Chinese fonning a sentence 'there is a mountain on top of another mountain', or 'there is a LlJ on top of another Lll', in reference to the graph 1±:1 which is more conventionally used to write ide- and which does look like one LlJ on top of another. The second, from MYS 11.2542, has nikuku (the infinitive of the adjective niku- 'hard') written by =}\. where is a regular ongana for !nil, but}\.+-, '8-1 01' here standing for a word meaning '81 ', is used for the syllables /kuku/ because the character for the word meaning 'nine' was used as ongana for the syllable /ku/ and 81 is equal to 9 x 9. The orthographic playfulness evident in these writings highlights that writing in Japanese in many cases was no practical matter of communication, but a leisure activity.

+-,

1.1.6

=

l?xar.nJ7les

The following are three examples of OJ writing, the first poems in the Kojiki and in the Man 'yoshii, respectively, and the first sentence in the first l?ngishiki

1.1 Writing

19

norito. The texts are transcribed using CAPITALS for logograms, plain italics for kungana and bold italics for ongana. (This is different from the transcription of OJ examples adopted through the rest of the book where we use italics for phonographic text and normal type for logographic text.) The Kojiki text is written entirely in ongana; it consistently uses the same kanji for each syllable, e.g. 4m for /tu/ in tatu 'rise', tuma 'wife', tukuru 'make'. The Man 'yoshii text is written in a complicated mixture of logographic and phonographic writing, using different ways of writing the same words, e.g. moti 'holding' written once phonographically as -BJ:~L, with an ongana and a kungana, and once logographically as~- The Norito text is written almost exclusively logographically, with only the complementizer to written in a small size ongana.

(5) a. Kojiki (KK 1) Text: ~?;\ ~~4fl){¥ lL ~~~~~4mWf.*1lilcfl~~~~4m?i\mtit1!g~~ ~~~ Interpretation: ~ V\~

~4m

-f¥R~ ~ ~ ~~

4fl)Wf. *1lilc 11

ya-kumwo eight-cloud

tatu rise.ADN

idumwo ya-pye-gaki Izumo eight-fold-fence

tuma-gomwi ni wife-enclosing DAT

~ ~ ~~ ~

~~~~

4m?i\mt

it 1!g

ya-pye-gaki eight-fold-fence

tukuru make.CONCL

so no ya-pye-gaki wo that GEN eight-fold-fence EXCL

'The many-fenced palace of Idumo Of the many clouds risingdwell there with my spouse Do I build a many-fenced palace: many-fenced palace!' (Philippi 1968:91) (NJ reading: yakumo tatsu Izumo yaegaki tsukuru sono yaegaki o)

To Ah, that

tsumagomi ni yaegaki

b. Man 'yoshii (MYS 1.1) Text: -~~~-£J:~L::ffJV\,~~~~~~$~t~.ffi7.J\~~~PJl*s~

.

~~if'~:fH:

Interpretation: ~~

KWO moyo miKWO moti pukusi mo yo mibukusi basket ETOP EMPH HON-basket hold.INF shovel ETOP EMPH HON-shovel ~

MOil hold.INF

20

Early writing in Japan and Old Japanese sources

Jlt KONO this GEN

*

IPYE home

~ NA greens

fir~ WOKAni hill DAT

~ ~J{

Jl

TUMA-su

KWO child

pick-RESP.ADN

s~~

~

kikana ask. OPT

NORA-sane tell-RESP.OPT

ifPtH:

'Girl with your basket, with your pretty basket, with your shovel, with your pretty shovel, picking greens on this hillside, I want to ask your home. Please tell me!' (NJ 'reading': ko mo yo miko mochi fukushi mo yo mibukushi mochi kono oka ni na tsumasu ko ie kikana norasane) c. Norito (EN 1) Text: ~fM';ft$ .:Et5L$ ~~ llf1 :i: ~ "§" Interpretation: • {';f t$.:£· UGWONAPAR-ERU KAMUNUSI, gather-STAT.ADN kamunusi

00-.i:

~

KIKOSI-MYESE hear.RESP-RESP .IMP

to COMP

tJL$ ~'

etr

PAPURI-RA, hafuri

MOROMORO all

NORU say.CONCL

'Hear me, all of you assembled kamunusi and hafuri. Thus I speak.' (Philippi 1990: 17) (NJ 'reading': ugonawareru kannusi, hafurira, moromoro kikoshimese to noru)

1.2

Sources

1.2.1

Japanese words inforeign sources

The Chinese history Wei zhi (M$ Chronicles of Wei, Japanese Gishi; a history of the Wei state (220-65), compiled towards the end of the third century) has a section, Wo ren chuan (~}..{];;Account of the dwarfs; Japanese Wajinden), which describes people living on the Japanese archipelago, also citing 53 phonographically transcribed words from the language spoken by these people. These words are often thought to be from an earlier stage of Japanese. Almost all are proper nouns or titles, with little known about their meaning. Among them are ;f~.~il' and -~Ill¥ which have entered Japanese folklore in the cNJ forms Yamatai (usually thought to represent an earlier form of Yamato (OJ

1.2 Sources

21

Yamato), an autochthonous name for Japan) and Himiko (the name of a mythical early female ruler which presumably would have been OJ *Pimikwo, but which has no OJ attestation). It is unlikely that the words in Wo ren chuan are anything but ad hoc transcriptions made by Chinese~ they should not be thought to represent conventional writings of these names and words. Indeed, the Japanese are not thought to have been literate in the third century. The transcriptions have not been deciphered, i.e. their sound values are not known, and they are not easily decipherable. The material against which they must be checked are the latest reconstructions of OC, which are, however, in themselves hypothetical. Even assuming sound values close to those which the characters had when used to write OJ words more than four centuries later, only in some cases do the words resemble known OJ words. Miller (1967: 1227) is an optimistic attempt to identify these transcriptions with Japanese vocabulary, representing the mainstream of traditional scholarship on the subject. It is not even evident that all of these words represent a language related to OJ. It is, in short, not possible to draw any conclusions about 'Japanese' on the basis of the transcriptions in Wo ren chuan. 1.2.2

Early inscriptions, wooden tablets and archival records

The oldest sources of Japanese in Japan are inscriptions made on stone and metal (swords and mirrors), the earliest thought to date back to the fifth century. The Japanese in these inscriptions is limited to proper names in texts otherwise written in Chinese or hentai kanbun. The following three are the most important early inscriptions: Inariyama tumulus sword inscription. ?471 (/?531). Excavated from a tumulus in Saitama prefecture. Thought to be the oldest inscription made in Japan. The inscription is dated as a cyclical year which corresponds to 471 ~ another possibility is 531. It is the longest early inscription, consisting of 115 characters, 46 of which are used phonographically to write Japanese personal and place names. The inscription employs features which are characteristic of early writing in Korea in terms of the shape of characters and the choice of characters used as phonograms~ it also contains a significant grammatical Koreanism (9=' used as a locative marker). Eda Funayama tumulus sword inscription. Late fifth or early sixth century. Kumamoto prefecture. The text is thought to have been seventy-five characters long, but because of lacunae only sixty-four are legible. It is in Chinese, but with the same grammatical Koreanism as the Inariyama tumulus sword inscription. It has Japanese names written phonographically. However, the writer is identified as iR:i( (Zhang An), someone of continental origin. Yakushi nyorai statue inscription. Probably second half of seventh century. Horyuji Temple in Nara. This is ninety characters long. It is the oldest extant

22

1 Early writing in Japan and Old Japanese sources

inscription written in hentai kanbun~ it has both some Japanese word order and some logographic expression of Japanese grammatical elements. There also exists a large corpus of more than 10,000 early archival records on paper (komonjo i]Jt:;=) from the eighth century~ out of these, however, only two are written phonographically in Japanese. Recently, a large number of wooden tablets with writing on them (mokkan Jl;::f'itl) dating from mid seventh to mid eighth century have been unearthed in different parts of Japan, the majority, however, around the old capital areas (Nara, Fujiwara). The existence of wooden tablets has been known for a while, but it is only in the post-war period, and particularly through the 1980s and 1990s, that large numbers have come to light. The total of these wooden tablets is now in excess of 150,000. They are usually short, often written in Japanese, with both logographic and phonographic writing represented. The writing ranges from labels through writing practice and scribbles to letters. Mokkan constitute the earliest evidence of popular writing in Japanese, showing that writing in Japanese was widespread in the second half of the seventh century and that literacy was not as confined as has previously been thought. Inscriptions, archival records and wooden tablets are valuable for their authenticity. They are, of course, enormously important to historians. They also provide important information about early use of writing in Japan and about the development of writing in Japanese. However, on the whole, these materials contribute relatively little to our understanding of the Old Japanese language as such when compared with the texts of the eighth century. It should be mentioned, though, that mokkan offer glimpses of language use which seems spontaneous and informal. However, the material is also very limited and mokkan may at best serve as a corrective to our knowledge of the Old Japanese gleaned from other sources.

1.2.3

Eighth-century texts

The bulk of our sources of OJ are texts from the eighth century. They have been handed down in copy and many competing manuscripts exist, all of which date from later periods. A long philological tradition, however, has made it possible to arrive at critical editions which are widely accepted as representing fairly well the texts of the eighth century. As mentioned, the texts comprise both logographic and phonographic writing. It goes without saying that it is the phonographically recorded texts or text portions which are the most important for the study of the language. It is important to keep in mind that no argument about the phonology or morphology of OJ can be based on the traditional readings of logographic

1.2 Sources

23

Table 1.1 Important OJ sources Kojiki (712) Harimafudoki (c. 715) Hitachifudoki (714-718) Nihon shoki (720) Izumo fudoki (733) Bungo fudoki (730s) Hizenfudoki (730s) Bussokuseki-ka (after 7 53) Man'y6shu (after 759) Senmyo (697-791) Kaky6 hyoshiki (772) Kogoshui (807) Engishiki norito (dates unknown; compiled 927)

passages. They can provide valuable clues in matters of syntax, but also in this area caution must be exercised. There are two main genres of text in the material: poetry and prose. Stylistically, these texts are either in highly formal, ritualistic prose or in a poetic form, ranging from folk-songs which were handed down (and thus subject to some editing both in that process and in that of recording) to elaborate poems. In any case, the language of these texts is probably in some aspects quite far removed from contemporary spontaneous and informal spoken language. In addition, there are Japanese vocabulary items and proper names in texts written in Chinese or in hentai kanbun, in the form of phonographically written items inserted directly into the texts, or explanatory notes written as part of the original text (as opposed to later additions). Needless to say, this in the main provides information about the OJ lexicon, not its grammar. Notes and glosses added onto Chinese texts in order to facilitate their interpretation and rendition into Japanese, the so-called kunten texts, constitute important material for the study of EMJ (see 6.2.2, 9.1.1). Although the practice probably had caught on already towards the end of the Nara period, surviving materials from that time are extremely scarce. Table 1.1 is a chronological list of main sources for the OJ period. 1.2.3.1 Poetry The main corpus for OJ is the poetry in the Kojiki, the Nihon shoki and the Man 'yoshu. This constitutes the material on which most of our knowledge

24

1 Early writing in Japan and Old Japanese sources

about OJ rests. The poetry comprises large portions written phonographically, thus making possible a comprehensive study of the phonology and morphology of OJ. Because of the general nature of poetry and the specific rhythmic constraints of Japanese poetry, with its dominant five- to seven-syllable metre, these materials provide less information about syntax. The Kojiki (i!-~2 'Record of ancient matters') is a history, compiled in 712. The preface is written in Chinese and the main text is written in hentai kanbun, but it also contains 112 phonographically written songs, as well as proper names and vocabulary. The Nihon shoki ( 13 ;js;::jUi2 'Chronicles of Japan') is also a history, compiled in 720. It is written in Chinese, but contains 128 songs, as well as proper names and vocabulary, written phonographically. TheManyoshu (77~~ 'Collection of myriad leaves') is the major source of the OJ language. It is a poetry anthology of more than 4,500 poems in 20 volumes. It was compiled in late Nara or early Heian. The latest poem is dated 759 and the earliest poem is usually said to date back to the middle of the fifth century. The poems are in Japanese, written both phonographically and logographically. Most poems present a mixture of phonographic and logographic writing, but the proportions differ enormously. The poems in the anthology are drawn from several sources, both older and contemporary. It is thus not a single text, but a collection of texts from different times, and consequently there is no overall orthographic consistency. There is, however, a substantial, identifiable portion representing early eighth-century OJ and another representing mid eighth-century OJ. Eastern OJ dialect is also represented in volumes 14 (azuma uta 'eastern songs') and 20 (sakimori uta 'border-guard songs').

1.2.3.2 Prose The prose corpus consists of two sets of texts: (a) Norito (:J:5G~jjj 'liturgies') comprise ritual prayers and blessings. Twenty-seven Norito are recorded in volume 8 of the Engishiki (JJf@:;J:\; 'Procedures of the Engi Era', completed 927) and are, despite the late date of compilation, thought to retain their OJ form and to reflect fairly accurately the OJ language. (b) Senmyo ("§"-i,J) are imperial edicts. Sixty-two edicts are recorded in the Shoku nihongi (~ 13 ;Js;: .*-2 797~ a history covering 697-791). BothNorito and Senmyo are written in senmyo-gaki, i.e., mainly logographically, but with grammatical items and some vocabulary noted phonographically. Although smaller than and very different from the corpus of poetry, these prose texts are extremely valuable for the study of some aspects of OJ syntax, especially the use of case particles. They also contribute to lexical studies and give an insight into ritual uses of language, as well as evincing clear evidence of influence from Chinese. The Norito in particular are thought to reflect quite early features of the language. Both Norito and Senmyo are as yet inexhaustively studied.

1.2 Sources

25

1.2.3.3 Others Fudoki (JOO.±~c 'Records of wind and earth') are topographies compiled on the order given in 713 by the empress Genmei to the governors of the provinces to provide information about local products and produce, the fertility of the land, and folklore. Five fudoki have survived: Hitachi fudoki (~~JOO.± ~(the north-eastern part of present-day Ibaraki prefecture), compiled between 714 and 718); Harimafudoki (ilmJOO.±~c (in Hyogo prefecture), compiled around 715), Bungo fudoki, Hizenfudoki (:!:~JOO.±~c (in Oita prefecture), HE!WJOO.±~c (in Saga and Nagasaki prefectures), compiled in the 730s), and Izumofudoki (l±l~JOO.±~c (in Shimane prefecture), completed 733), which is the only fudoki to have survived in its entirety and which is written in hentai kanbun, whereas the rest are written in Chinese. They contain phonographically written poems (some twenty), place names and local vocabulary. The Bussokuseki-ka ({.L.JE.:t:i;JX 'Footprints of the Buddha poems') is a series of twenty-one poems inscribed on a stone at the Yakushi-dera in Nara. The provenance is unknown, as is the date of inscription, but it is thought to date after 753. The poems are written entirely phonographically and are valuable because of their authenticity as a primary source. The Kakyo hyoshiki (;£}\)jl}f~~) is the earliest known poetic treatise written in Japan, from 772 and by Fujiwara no Hamanari. It is written in Chinese, but has phonographically written poems in Japanese. The Kogoshiii (i!~~J;: 'Gleanings in Old Words') is a history, compiled in 807 by Imbe no Hironari (then in his seventies). It is mostly in Chinese, but contains two phonographically written poems and some vocabulary. REFERENCES Writing in Korea: Solm 1999: 124--8. Early Korean-Japanese contact Lewin 1976, Bentley 2000. Introduction of writing and adaptation of Chinese script in Japan: Miyake 2003a, Seeley 1991. Man yogana: Lange 1973, Miyake 2003a. Overview over sources: Vovin 2005: 1-20. Inscriptions: Murayama and Miller 1979, Seeley 1991: 16-40. Reproductions and transcriptions of most inscriptions: Kariya et al. 1912. Wooden tablets: Farris 1998: 201-32, Kotani 1986. Wei zhi: Farris 1998: 10-14, Miller 1967: 12-27, Tsunoda and Goodrich 1951. Norito and Senmyo: Bentley 200la.

2

Phonology

Early Old Japanese had the distinct syllables shown in Table 2.1, illustrated with one representative man 'yogana for each syllable.

2.1

KO-rui and otsu-rui syllables

OJ kept distinct so-called ko-rui (type A) and otsu-rui (type B) syllables, which merged in the transition to EMJ. The difference is here noted by subscript '1' and '2'. Thus, the two distinct OJ syllables represented by, amongst others, 3'E and* respectively are termed me1 (ko-rui me) and me2 (otsu-rui me) because they merged and are reflected as EMJ (and NJ) me. The following are examples of minimal pairs. (1)

OJ

EMJ

NJ

pil pi2

'sun' 'fire'

pi

hi

me1 me2

'woman'} 'eye'

me

me

ko1 ko2

'child' 'this'

ko

ko

Phonemically, the difference between the ko and otsu syllables is generally agreed to pertain to the post-consonantal part of each syllable. However, despite the subscript convention used here and elsewhere, and despite the way the orthographic ko-otsu distinction is often talked about, it is important to appreciate that OJ did not have 'two kinds of' Iii, lei or lol, any more than Classical Greek had many kinds of 'i'. EMJ Iii, lei and lol each represent the outcome of a merger between two distinct sounds of OJ, just like Modern 26

Table 2.1 Orthographically distinct syllables in early OJ ~iiJ

.a

:IJO ka

'IX ga

X: ki,

*.ff; gi,

{JT.i

~

.u

tx. .e

:i!i:ki,

~gi,

0- ku

,!!. gu

*ke1

;;f ge,

mke,

age,

15" ko 1

~go,

ii:. sa

tl" za

~ta

~t da

~na

Zsi

§ zi

~ti

~di

'itni

~su

~zu

:ml tu

lidu

ft.)( nu

~se

:lik ze

~te

S3 de

ft!. ne

it so,

m.zo 1

JJ to 1

IJ!t do 1

j!; go,

1t so,

fD. zo,

.lt to,

!1\f do,

~ba

U*ma

.l:t pi,

Jif.bi,

~mi,

n~

pi,

{lllj bi,

*mi,

;(ji pu

.*bu

.$mu

!J!. pe,

#be,

~me,

J=ipe,

f* be,

*me2

&"po

~bo

~no,

~.0 ~ko 2

781: pa

l'J no,

~ya

.Bt ra

fUwa

flj ri

~wi

S:!yu

l5ftru

}.!!; ye

1L re

=f:.mo1

ffl yo,

nro,

-amo,

#;yo,

8 ro,

~we

.lJl.wo

28

2 OJ phonology

Greek /i/ reflects the merger of a num her of distinct sounds of Classical Greek (including u /y:/, st /e:l, 11 /d, ut /yil, T]t IE:i/, ot /oil). The distinction between the ko- and otsu-rui syllables disappeared from the language in the transition to HMJ and is not reflected in the kana writing of EMJ or later periods. As each OJ syllable could be represented by different man 'yogana, the distinction was not conspicuous to later generations of readers and in fact it was only finally discovered at the beginning of the twentieth century by Hashimoto Shinkichi, one of the great Japanese linguists and philologists. This is Hashimoto's own account of his discovery (c. 1915, cited from Ohno 1980: 139ff.; my translation). While in February 1909 I was researching developments in the syntax of Japanese at the request of the Japanese Language Investigative Committee [Kokugo Chosa linkai], I noticed in the eastern songs [azuma uta] in volume 14 of the Man'yoshU that the character * in not a few cases was written in places which ought to have the particle ga. This gave rise to the suspicion that the eastern dialect of that time had a particle ke which was used in the same way as ga. As a way of solving this I felt it necessary to check up on every single*· First I examined volume 14, but to no avail. Because there are examples in volume 5 of * used in the meaning of ga I decided also to examine volume 5. This time I did not limit myself to the character *' but collected every single kana for ke. I then foWld- and this had no bearing on the problem I was investigating - that forms such as ni-keri 'perfective-modal.past', ke-mu 'pastconjectural', kerasi 'past. presumptive', and kepu 'today' were written exclusively with characters from one set, comprising e.g. * and $t, whereas words such as take 'bamboo', sake 'sake', and take 'moWltain, peak', nageki 'sigh', sigesi 'thick, dense (of growths)' were written exclusively with characters from another set, comprising e.g. ~ (ke) and~ (ke), 1l: (ge). I discovered that among words written with ke some characters could be used interchangeably, while others could not, and that kana for ke accordingly fell in two groups; the distinction between these two groups was maintained consistently. I felt strange discovering this. Experiencing great interest I decided to proceed and examine the other volumes. I wanted to begin with the eastern dialect and examined volume 20, but in the eastern dialect border-guard songs [sakimori uta] I was not able to verify this distinction, which made me feel greatly disappointed. However, in the poems in the central dialect in that volume, I found that this distinction clearly existed, and I thus came to the conclusion that this was a distinction which existed in the central dialect, but not in the eastern dialect. Next, I examined those volumes of the Man 'yoshu which are written almost entirely phonographically, volumes 15, 17, 18. There would occasionally be a few exceptions, but I foWld that in almost all instances this distinction was there. When I looked at the remaining volumes of the Man 'yoshU as well as the songs in Nihon shoki and Kojiki, apart from volume 14 of the Man 'yoshU, the distinction between these two types was there, without a single exception. This led me to believe that this distinction generally held for the Nara period- excepting the eastern dialect. Going on to check the exceptions I had previously foWld in the Man 'yoshU against an old copy of the Man 'yoshU in the possession of Oya T5ru [1850-1928], I learned that almost all were not real exceptions, but mistakes in the wood-block printed editions, and I arrived at an even

2.1 Ko-rui and otsu-rni syllables

29

firmer belief in the existence of this distinction. (In the following year, 1910, I was able to take a look at a Genryaku-period (1184) manuscript of the Man 'yoshi:i, which agrees with Oya's. It was increasingly clear that the printed editions were mistaken.) When I next examined texts from the Nara period such as Jogil Shotoku-hoo taisetsu [late seventh century], texts in the Dai-Nihon komonjo [a compilation of handwritten materials, published 1901-40 by Tokyo University Press] and the Senmyo in the Shoku nihongi there was not a single exception. I wanted to look at the traces of the changes in this kana-use and therefore examined texts from the Heian period written in man 'yogana such as edicts and songs contained in the official histories from Nihon koki [840] onwards, Nihongi kyoenka [compilation from three occasions (882, 906, 943) of poems composed at the court], Shinsenjikyo [a dictionary, compiled c. 898901 ], and the Nih on ryoiki [a collection of tales, written mainly in kanbun, completed after 822]. In most it was not possible to observe this distinction. Thus, I was able to learn that this distinction between two types of ongana for ke existed in the Nara period, but had broken down in the Heian period. While pursuing the investigation of ongana ke, I also examined the k-initial syllables in volume 5 of the Man 'yoshi:i in order to check if there was such a distinction for other kana. I was able to infer that besides ke there must have been also two types of ki and ko. I had previously gathered all the verbs in the Nihon shoki and Kojiki according to inflected forms. Looking at these in the light of my research on ke, I was furthermore able to learn that there must have been two kinds of pi, mi, pe, and me. This kind of kana-use was something no one else had hitherto explained, and just when I thought that this was one of the few original discoveries since the time of Keichfi [famous Edo period philologist, 1640-1701] and had decided to proceed with my investigation of each kana in la~ger materials, I accidentally looked at Kogen betsu 'on sho [c. 1849, by Kusakado Nobutaka, 1818-69] which had just been acquired by the Japanese Language Department at Tokyo Imperial University. Thereby I learnt that this kind of research had been done by someone else previously, and when I looked at the Kana-zukai no Oku no yamamichi [1798, by lshizuka Tatsumaro, 1764-1823] on which that book is based, I realized that it shows that there was a distinction between two types of kana on a much grander scale than what I had researched and been able to infer. My discovery was really a rediscovery. Thus, this discovery of mine was a discovery in a two-fold sense. One was the rediscovery of this special kana-use, the other was the discovery of lshizuka Tatsumaro's unknown studies into kana-use. If I had not on my own discovered this kana-use, I surely would not have been able to understand the importance of Kana-zukai no Oku no Yamamichi, or to appreciate its value. Thus, only because I had investigated this independently, I grasped the merits of the studies of that man of old. Ever since Hashimoto's discovery of this orthographic distinction, the linguistic interpretation of it has been the dominant topic of research in Old Japanese phonology. It is a matter which continues to attract attention and different views. As there is no convenient way of representing the distinction in the present-day Japanese writing system, it remains widely ignored outside linguistics. This is probably to some extent a consequence of the lack of consensus about its phonemic interpretation, but it is regrettable and makes difficult a full appreciation of the sound texture of OJ poetry.

30

2 OJ phonology

As shown in Table 2.1, the earliest attested Old Japanese distinguished eighty-eight syllables. The distinction between mo1 and mo2 is found in the Kojiki (712), but had merged in later sources. Sometimes a distinction is also posited between po1 =F po2 and bo1 =F bo2 . It is questionable whether the existence of such an orthographic difference can be demonstrated within the OJ text corpus, but it is almost certain that *po1 =F *po2 and *bo1 =F *bo2 were phonologically distinct in slightly earlier Japanese. The syllable inventory in Table 2.1 represents a system in change and the gradual breakdown of the distinctions is evident through the OJ text corpus. The last syllable pair to be distinguished was ko 1 =F ko 2, which were kept distinct into the early tenth century, but by then probably merely as an orthographic convention.

2.1.1

Co 1 versus Co2

There are not many minimal pairs distinguished by the difference between Co 1 and Co2 and it has been proposed (Matsumoto 1984) that Co 1 and Co2 in fact did not represent distinct syllables, but were allographic variants, possibly standing for allophonic variants. This is a consequence of the following distributional differences between Co 1 and Coi Most occurrences of Co1 are in morpheme final position and Co 2 did not occur in a root morpheme with Ca, Cu, Co 1, see below. Also, there are not many monosyllabic lexical morphemes with Co 2 . However, while the scope for opposition is fairly limited, there are minimal pairs, e.g. ko 1 'child' versus ko2 'this', making clear that the two sets of graphs did stand for distinct syllables.

2.1.2

Phonetic reconstruction and phonemic interpretation

We face two distinct, but interrelated, problems concerning the phonological understanding of the OJ syllable inventory: phonetic reconstruction and phonemic interpretation, in particular segmentation. Phonetic reconstruction is based mainly on the following two types of evidence: (a) external: comparison with the phonetics of other languages, first of all comparison with the (Early Middle) Chinese sound values of the kanji used as on-gana; (b) internal: comparison with the phonetics of later stages and dialects of Japanese, and sound changes within Japanese, either documented changes in EMJ (or later) or reconstructed changes in pre-OJ. Such internal and external evidence can provide valuable clues, but it also has important difficulties of interpretation and includes a number of unknown factors. Arguments based on internal evidence run the evident risk of circularity. Regarding external evidence, as mentioned above (1.1.2.5), the choice of kanji to act as phonograms was not exclusively based on phonological considerations; furthermore, the actual relation between EMC and OJ is anything but direct, as Chinese came to Japan

2.1 Ko-rui and otsu-rui syllables

31

by way of the Korean peninsula and was very likely based on a different stage, if not variety, of Chinese; and finally, EMC is itself a reconstruct, i.e. hypothetical. For this type of reconstruction, Miyake (2003a) is an extremely useful and well-documented study with an impressive command of all the relevant materials.

2.1. 3

Sound values

Miyake's broad reconstruction posits the basic sound values shown in (2). Some of these sound values are more similar to what we reconstruct for pJ (2.7.2) than for OJ, but overall they are compatible with the phonemic interpretation adopted below. (2)

il i2 e1 e2 a 01 02 u

[i] [i]

[e] [;)y] [a] [o]

[;)] [u]

More narrowly, the syllable pairs seem to have differed phonetically as follows: (3)

Ci1 more palatal Ci2 more labial;

[Cii] [Cl)i, Cui]

Examples pil [pii] 'sun' pi2 [pl)i, pui] 'fire'

[Cei, Cai] [Cie]

me2 [meL mai] 'eye' me1 [mie] 'woman'

[Co] [Cl)o]

ko2 ko1

a falling diphthong

Ce2 a falling diphthong Ce 1 more palatal; a rising diphthong

Co2 a monophthong Co1 more labial;

[ko] 'this' [kl}o] 'child'

a rising diphthong

2.1.4

Phonemic interpretation

The syllabic writing gives no clues regarding the organization into segments of distinctive phonic qualities and that makes the phonemic interpretation difficult. Even so, the difference between the ko and otsu syllables is, as

32

2 OJ phonology

Table 2.2 Transcription systems for OJ

Syllable type

Index notation

/(jj-rui Otsu-rui

i, i, i

neutral

Ohno

Modified Mathias-Miller

Yale

Frellesvig & Whitman

yi iy

wi

ye ey e

ye e e

/(jj-rui Otsu-rui

e,

e

e

~

e

e

neutral

e

e

e

/(jj-rui Otsu-rui

o, o,

0

0

WO

WO

0

0

Q

0

neutral

0

0

0

0

0

u

u

u

u

u

a

a

a

a

a

mentioned above, generally agreed to pertain to the post-consonantal part of each syllable. Traditionally, the difference was thought to be a distinction in vowel quality, leading to the hypothesis that OJ had eight distinct vowel phonemes, with many different proposals concerning the structural organization of these eight vowels. However, since Lange (1973), it is common to interpret the difference in terms of sequential diphthongs, i.e. as being due to the presence of a palatal or labial glide, /y/ or /w/, in one or the other mem her of the syllable pairs. The phonemic interpretation and transcription which will be adopted here is shown in Table 2.2, based on Frellesvig and Whitman (2008a). For reference, the Yale system of transcription (in Martin 1987), Ohno's system (1990), and modified Mathias-Miller are also included. Table 2.3 gives some examples. On this interpretation, Old Japanese had the five vowels also found in Middle and Modern Japanese: li, e, a, o, u/. In addition to identifying the properties distinguishing ko and otsu syllables, the assignment of neutral syllables, i.e. syllables with undistinguished Ci, Ce, or Co, is an issue. As shown, they are here phonemically identified with Cij, Ce2 , and Co2 , respectively. Table 2.4 gives the inventory of distinct syllables in this phonemic notation. All diphthongal interpretations suffer from some awkwardness from a structural point of view in the distribution of glides, mainly because they usually incorporate diachronic considerations. The one adopted here has the advantage of positing only sequential diphthongs that are also found as free syllables. For some of the syllables involving the labial glide, in particular /mwo, nwo, ywo, rwo~ pwi, bwi, mwi/, it seems reasonable to assume a pronunciation like [uQ] and [ui], that is, with a shift in phonetic sonority peak from the vowel to the glide.

2.1 Ko-rui and otsu-rni syllables

33

Table 2.3 Examples of transcribed OJ forms

Gloss

cNJ

Frellesvig & Whitman

Index notation

Yale

Modified Mathias-Miller

Ohno

'sun' 'fire' 'blood'

hi hi chi

pi pwi ti

pi, pi, ti

pyi piy ti

pi pi ti

pi pi ti

'woman' 'eye' 'hand'

me me te

mye me te

me, me, te

mye mey te

me me te

me me te

'child' 'this' 'ear (of rice)'

ko ko ho

kwo ko po

ko, ko, po

kwo kQ po

ko ko po

ko ko po

Table 2.4 Phonemic syllables in OJ .a .i .u .e .0

ka ki kwi ku ke kye ko kwo

2.1.5

ga gi gwi gu ge gye go gwo

sa si

za

su se

zu

so swo

ta ti

da di

na ni

ze

tu te

du de

nu ne

zo zwo

to two

do dwo

no nwo

zi

pa pi pwi pu pe pye po

ba bi bwi bu be bye bo

rna mi mwi mu me mye mo mwo

ya

ra ri

wa wi

yu

ru

ye

re

we

yo ywo

ro rwo

wo

Neutralization

It is notable that -i and -wi, and -ye and -e only were kept distinct after /p, b, m, k, g/; after glides and alveo-dental consonants the distinction was neutralized, cf. the imperative and exclamatory forms of consonant base verbs in (4) and the gerund forms of quadrigrade (5a) and upper bigrade (5b) verbs: (4)

Base kak- 'write' mat- 'wait'

(5) a.

b.

Imperative /kakye/ //matye// =>/mate/

Base ok- 'put' or- 'weave'

Gerund /okite/ /orite/

ori- 'descend' okwi- 'arise'

/orite/ /okwite/

Exclamatory /kake/ /mate/

34

2 OJ phonology

It is likely that there earlier was a distinction li/ ::f /wi/ and lye! ::f lei after all true consonants and that the state exhibited by OJ reflects a system in collapse. On the other hand, /p, b, m, k, g/ constitute the class of grave consonants and the environment of distinction is therefore general. It is thus also quite possible that the OJ state reflects a stable stage of some standing. The distinction between /mwo/ and /mol is thought to have been found only in the Kojiki, although some scholars posit it for portions of other texts. In the later sources /Cwo/ and !Col were only kept distinct where C was not a labial consonant. This, however, does not reflect systematic neutralization, but simply the short-term course of the merger of the distinction. Thus we assume that there was distinction between /*pwo/ ::f /*pol and /*bwo/ ::f !*bo/ in slightly earlier Japanese. 2.2

Consonants

OJ had the following inventory of consonant phonemes. The mediae (/b, d, g, zl) and the liquid (/r/) did not occur word initially, see below. There is nothing remarkable about the phonetics and phonology of the nasals, glides, and the liquid. The liquid was a flap, like its NJ reflex: [r ], in the following simply noted by [r]. Note that the palatal glide /y/will be noted by IPA [j] in phonetic transcription. (6) Tenuis Media Nasal Liquid Glide

2.2.1

Labial

Alveo-dental

p b

t, s d,z n

m

Palatal

Velar k g

r

w

y

Obstruents

The obstruents, on the other hand, exhibited several interesting features. The main phonetic variants may be thought to have been as in Table 2.5, noted in a broad transcription. Phonologically, the phonetic system manifested in Table 2.5 may be understood in terms of four overlapping sets of obstruents: tenuis versus media, and sibilant versus non-sibilant obstruent, which were distinguished by two distinctive feature categories, tenseness (+/-tense) and stridency (+/-strident), and which displayed phonetic variation with regard to voicing, nasality, and continuousness.

2.2.2

Tenues versus mediae; medial voicing and prenasalization

The traditional Japanese terms for tenues and mediae are sei 'on O'il!if 'clear sounds') and daku 'on (~if 'muddy sounds'), respectively. The mediae

2.2 Consonants

35

Table 2.5 Phonetic realization of OJ obstruents Phonetic realization Phoneme

Word initial

Word medial

Phonological classification

/p/ It/ /k/

[p-(1 [t [k- X

b -13] d] g -y]

non-sibilant

lsi

[s- 's

z-"z]

lbl /d/ lgl

["b- "13] ["d] [•g -•y]

lzl

["z- ""z]

}

sibilant

}

non-sibilant sibilant

}

tenuis (sei 'on)

}

media (daku'on)

(lb, d, g, zl) were phonetically prenasalized. That is to say, they had a nasal onset or onglide. This feature of pronunciation is generally thought not to have been lost until early in the NJ period. We will refer to this simply as prenasalization. Evidence comes both from EMC sound values for the ongana for syllables with initial media and from a number of sound changes in which nasals and mediae behave alike, e.g. the onbin sound changes which took place in the transition from OJ to EMJ (see 7.1.4), cf. OJ yomite 'read.GER' > EMJ yoNde, OJ ywobite 'call.GER' > EMJ yoNde. Note in this connection that to all appearances vowels before nasals and mediae were allophonically nasalized. The tenues (/p, t, k, sf) were distinguished from the mediae in terms of the phonetic parameter of tenseness: the tenues being tense and the mediae lax. The tenues were allophonically voiced in word medial position. In a narrow phonetic transcription this might be noted [p, t, ~, §],but for the sake of convenience [b, d, g, z] will be used here. Voicing, then, was subject to variation and was assigned by a redundancy rule; it was not a distinctive phonetic feature in OJ. This is the reason that we prefer to speak of /p, t, k, s/ as tenues and of /b, d, g, zl as mediae, rather than as unvoiced and voiced obstruents. Voicing of tense obstruents in word medial position will be referred to in the following simply as medial voicing. The hypothesis of medial voicing is due to Wenck (1959). Evidence for it comes mainly from sound changes, first of all 1-p-1 [b- .(3] merging with 1-w-1 in EMJ (see 7.3.1.1), secondarily from the kind of syllable weakening involved in the onbin sound changes. Thus, the distinction between tense and lax obstruents in word medial position was phonetically manifested primarily as one of prenasalization, cf. pata 'flag',pada 'skin', andpana 'flower', (7):

36

2 OJ phonology

(7)

Tenuis /pata [pada

Media pad a panda

Nasal pan a/

pana]

Similar phonetic realizations oftenuis and media are retained in some dialects (for example in northern Japan or in Shikoku), but mostly prenasalization and medial voicing were lost in the course ofLMJ (see 11.1). That phonetic change led to major differences in sound texture between OJ and NJ. We illustrate these differences in (8), where we also assume that OJ word initial tenues were not aspirated, or at least not as aspirated as NJ tenues. The word /tanabata/ 'Vega, the Weaver' has not changed phonemically over more than a millenium between OJ and NJ, but its phonetics have changed considerably. Because of prenasalization and medial voicing, OJ sounded very different from NJ. (8) OJ NJ

2.2.3

/tanabata/ [tanambada] [fanabafa]

Non-sibilant versus sibilant obstruents

The sibilants (/s, zl) were pronounced with a sibilant (high intensity) noise and were distinguished thereby from the non-sibilant obstruents (/p, t, k, b, d, g/) which were pronounced with no such noise. However, both sibilant and nonsibilant obstruents exhibited variation with regard to continuousness. That is to say, both continuant (fricative) and abrupt (occlusive and affricative) sound types are found among both classes of phonemes. This is the reason we prefer to speak of /s, zl versus /p, t, k, b, d, g/ as sibilant versus non-sibilant obstruents, rather than as fricatives versus stops. The sibilants had both fricative, [s-, -z; _nz], and affricated, [ts-, _dz; _ndz], variants. The distribution seems to have been conditioned by the following vowel (here following Kobayashi 1981 and extending her findings to /-s-, -z-1): (9) /-a/ 1-o/ 1-u/ 1-wo/ /-i/ /-e/

Word initial /s-

Word medial -z-1 -s-

[tsa] [tso] [su -1;u] [swo] [Ji] [Je]

[dza] [dzo] [zu- dzu] [zwo] [3i] [3e]

[ndza] [ndzo] [nzu- ndzu] [nzwo] [~i] [~e]

2.2 Consonants

37

The phonetic reconstruction of /s, zJ remains debated, with proposals ranging from fricatives only (e.g. Sandness 1986, Miyake 2003a) to affricates only (e.g. Ogura 1998). If any, the consensus view is probably the one adopted here with variation between [1s] and [s]. The debate is to some extent obscured by a lack of clear distinction between phonetic and phonological reconstruction. It should be noted that common to all proposals, though not explicitly addressed, is that /s, zl are manifested as sibilant sound types. The evidence concerning the reconstruction of /s, zl comes from two main sources: (a) Reconstructed EMC sound values for the ongana for syllables with initial /s, zJ. Miyake's material clearly shows that among the ongana for sa, so, and su, many, some, and a few, respectively, are reconstructed with initial fricative. For zV, many of the ongana are reconstructed with initial affricate, though not as neatly distributed as for sV (b) The priest Ennin's Zaitoki (858), in which he described Sanskrit sounds by means of illustration with phonograms used for their Japanese or Chinese sound values. Ennin seems to equate the pronunciation of Sanskrit ca ([tJa]) with Japanese sa and that is often taken as evidence for an affricative sound value for /s/, at least before /a/. Apart from the fact that Zaitoki primarily concerns EMJ and not OJ, the interpretation of the text is far from straightforward, as pointed out by e.g. Sandness (1986). Finally, Ainu casi [tJasi] borrowed from OJ sasi 'castle' (in turn borrowed from a Korean language, reflected as MK cas) also points towards an affricative pronunciation of OJ /s/, at least before /a/. /p/ had both occlusive (abrupt), [p-, -b], and fricative (continuant), [p-, -13] variants. It is not possible to identify any conditioning phonological environment for this variation, so it is possible that [p-, -b] and [p-, -13] may have been in free variation. Earlier it was thought that /p/ had already shifted completely to [p] in OJ, but there is little to support that view. It seems that similar variation was displayed by other non-sibilant obstruents; in particular lk, b, g/ seem to have had both abrupt and continuant variants. Some of this variation was likely subject to stylistic variation, with abrupt variants being more characteristic of careful and continuant variants more characteristic of casual speech. Evidence for continuousness variation for the non-sibilant obstruents comes from sound changes, first of all /-p-/ [b - 13] merging with 1-w-1 in EMJ (7.3.1.1), secondarily from the kind of syllable weakening involved in the onbin sound changes (7.1.4). In this way, continuousness was subject to variation and both of its values were represented in the realization of members of both sibilant and nonsibilant obstruents. This phonetic variation will be referred to as continuousness variation.

38

2 OJ phonology

(10) /p- t- k[p t k [ip (x)

Abrupt Continuant

2.3

Non-sibilant -p- -t- -k- -b- -d- -gb d g mb nd Dg (y) cmp) (Dy) _[3

Sibilant s- -s- -z-1 ts dz ndz] nz] s z

Other allophonic variation

Two fairly low-level phonetic allophonic rules may be posited, (11) and (12), exemplified in (13). Nasalization probably also to some smaller extent applied to vowels after nasals. (11)

Nasalization of vowels: Vowels before nasals or mediae were phonetically nasalized (reflected for example in a variant spelling of uma 'horse' as .$.~ 'mu.ma').

(12)

Tonality a4fustment: Consonants before li, y/ were phonetically palatalized (sharpened), and consonants before /u, w/ were labialized (flattened).

(13)

uma 'horse' [fun a

kimi 'lord' kiimii

ywobi 'calling' jl}Ombji

kubi 'head'

pikye 'pull!,

kwU~ji

Piigje]

Palatalization probably also to some extent applied to consonants before /e/, see (14). That is to say, 'neutral /Ce/' was phonemically like Ce2 , but phonetically more like Ce 1 . Presumably, the gradual emergence of uniform phonetic palatalization before /e/ is what led to the disappearance of a distinctive phonemic glide in Ce 1 (!Cyel) syllables. me1 'woman' te 'hand' me2 'eye'

(14)

Ce 1 Ce Ce2

2.4

Phonetic transcription of a text

/Cye/ /myel ICe/ /tel ICe/ !mel

[Cje] [Cie] [Ce]

[mje] [tie] [me]

On this background it is now possible to illustrate what OJ probably sounded like, giving a broad, but still reasonably detailed, phonetic transcription of a short text (the first poem in the Kojiki, cf. 1.1.6), with a. illustrating more careful and b. more casual diction.

2.6 Morphophonemics

~0.~

(15) a. b.

a. b.

$4m

/yakumwo tatu [jagwfunUQ tadwu [jaywfunu9 tadwu

39

{ti-ll.~ ~m~~

4mM*1lilc fl

idumwo yapyegaki i"dwfunu9 jabjeugag} i"dwfunu9 ja[3jeuyay}

tumagomwi ni twfunaugomui nii twfunauyomui ~i

~m~~

4m0.¥m

it

yapyegaki jabjeugag) ja[3jeuyay)

tukuru twugwuru twuywuru

so no tso no so no

Ab

!'It:.

~m~~ ~

yapyegaki wo/ jabjeugagii wo] ja[3jeuyayii wo]

'The many-fenced palace of Idumo Of the many clouds rising - To dwell there with my spouse Do I build a many-fenced palace: Ah, that many-fenced palace!' (Philippi 1968: 91)

2.5

Syllable and word structure

As seen in Table 2.4 above, OJ syllables have the following structure: (C)(G) V. Syllables consisting of a single vowel are generally restricted to word initial position, except for a few exceptions, including a single noun, ka.i, and inflected forms in the bigrade conjugations ( cf. 3.4.1.2), such as ku.i 'regret. INF'(<= 1/kuyil!, cf. conclusive kuyu) and ku.u (<= /!kuwu/1, cf. infinitive kuwe-), i.e. verbs whose bases end in a syllable with initial glide. Thus all syllables are open and short, surface forms having to conform to repetition of the simple (C)(G)V pattern. A number of distributional facts about OJ sounds are best understood in diachronic terms and will be explained in the section on proto-Japanese: (a) absence of word initial mediae or liquid (2.7.1.2)~ (b)Arisaka's Law (2.7.2.1)~ (c) limited distribution of Cwo, Cye, Ce, and Cwi (2.7.2.4)~ (d) noun apophony (2.7.2.2).

2.6

Morphophonemics

2.6.1

Vowel deletion

When in morphological derivation or word formation two vowels come together without an intervening consonant, one is elided. Depending on the number of syllables (one or more) in the two morphemes, the following rules apply. The second vowel is only elided when a monosyllabic morpheme is followed by a vowel initial polysyllabic morpheme, i.e. (16a)~ elsewhere the first vowel is elided (16b):

40

2 OJ phonology

(16)

Synchronic rules of vowel deletion a.

b.

V1 +V2 =>V1 => /-CV 1 + V 2CVI wa + ga + ipye 'I+ GEN +house'=>

/-CV 1CVI wagapye 'my house'

vl + v2 => v2 elsewhere /CVCV1 + Vz/ ake + u 'redden+ CONCL'

=> =>

/CVCVzf aku. 'redden.CONCL'

/CVCV1 + V 2CVI myesi +age 'see.RESP +raise' waga + ipye 'my+ house'

=> => =>

/CVCV2CVI myesage- 'summon' wagipye 'my house'

+ ipa 'eternal + rock'

=>

tokipa 'eternal ever lasting'

=> =>

/CVz/ ki 'come.INF'

toko

/CV1 + Vzf ko + i 'come + INF'

rock~

It should be noted that there are apparent counterexamples to (16a), such as the competing form wagipye in (b). However, in such cases a different constituent structure may be assumed for the underlying form, here waga-ipye with univerbation between pronoun and genitive particle. Another similar set of forms are those where the existential verb ar- seemingly fuses with a preceding grammatical monosyllabic morpheme: the periphrastic stative -te-argiving -tar-, the extended negative -(a)zu-ar- giving -(a)zar-, and the extended adjectival copula -ku.-ar- giving -kar-. In these cases, however, ar- must be thought to fuse with the full inflected form, not just the suffix: kakite-ar- => kakitar-, kakazu-ar- => kakazar-, akaku.-ar- => akakar-. In OJ all such forms were simple phonological fusions (3.4.2.1.2), but in EMJ they were reanalysed to give rise to the morphemes -tar-, -zar-, and -kar- (8.2.1). Similarly, the EMJ inflected copula nar- seems to be from ni-ar-, but also here an intermediary derivational step must be posited: iku.sa-ni ar- => ikusani-ar- => iku.sanar(ikusa 'army'). The same holds for the copula tar- which is found from EMJ and which is formed from the copula infinitive to and ar-.

2.6.2

Rendaku

In the morphophonemic process known as 'rendaku', an initial tenuis in the second component of a compound changed to a media to express close compounding. 'Rendaku' is usually rendered 'sequential voicing' in English, but as voicing was not the phonetic feature which distinguished tenues from mediae, that is an inappropriate translation, so here 'rendaku' will simply be

41

2.7 Proto-Japanese

retained. It is thought that this process originated in the reduction of a particle with initial nasal, usually identified with OJ genitive no or dative ni. Thus sakurabana is thought etymologically to derive from sakura-no-pana, or yamadori from yama-no-tori. However, examples such as (18) show that rendaku already in OJ was established as a morphophonemic process, for they cannot be etymologized with no or ni: na 'you' invariably takes ga as genitive particle, never no, and yomwi ni kapyer- would mean 'return to Hades', not from. Examples such as tuma-gomwi 'wife-enclosing' and tuma-gwopwi 'wifeloving' incorporate a direct object, again not directly derivable from no or ni or any other n-initial particle. (17)

sakura 'cherry'+ pana 'flower' yama 'mountain'+ tori 'bird'

=> sakurabana 'cherry-flower' => yamadori 'copper-pheasant'

(18)

=> nadori 'your bird' na 'you' + tori 'bird' yomwi 'Hades'+ kapyer- 'return' => yomwigapyer- 'revive (intr.)'

As forms resulting from rendaku were lexicalized, a number of alternations resulted, shown in (19), some of which are not phonologically automatic because of later sound changes (/p/ >If/> /hi (see 11.3, 14.3) and /d/ > /zl before li, u/ (see 14.1). (19)

p- > LMJ f > NJ htt-

ks-

-b-d-d- > NJ -z- (/ _ {i, u}) -g-z-

Rendaku is not phonologically predictable. It is, however, blocked when the second component contains an internal media, e.g. kamu- 'spirit'+ kaze 'wind' => kamukaze 'divine wind', not *kamugaze. This regularity is known as Lyman :SLaw, named after B. S. Lyman (1894, see Vance 2007). It continues to hold in NJ where rendaku is still very active.

2. 7

Proto-Japanese

It is possible to reconstruct a simple pJ phoneme inventory and some of the sound changes that took place between pJ and OJ. The main point of interest is that some phonemes and diphthongs of OJ arose through contraction of sequences of segments in pre-OJ. Table 2.6 gives an overview.

42

2 OJ phonology Table 2.6 Phonemic Con-espondences between OJ and pJ Primary OJ consonants

pJ

p

*p *t *k *s *m *n *r *w *y

I

k s m n r

w y Secondary OJ consonants

Pre-OJ sequence

b d g

*mVp, *nVp; *np *mVt, *nVt; *nt *mVk, *nVk; *nk *mVs, *nVs; *ns

z Primary OJ vowels and diphthongs

pJ

u

*i *e *a *o *u

0

*i

0

*~

i, ye

a U, WO

2. 7.1

Secondary OJ vowels and diphthongs

Pre-OJ sequence

wi

*uy, *iy

e

*ay,

ye (wo

*ia (*i~) *uo, *ua)

*~y

Consonants

2. 7.1.1 Secondary origin of OJ mediae (/b, d, g, z/) The OJ mediae reflect contractions of a nasal with a following obstruent. Where recoverable, such sequences arose through weakening and loss of a syllable, often where univerbation of a complex form took place; this was also the origin of rendaku (2.6.2). It should be noted that the syllable loss involved is sporadic, that is, not part of regular sound changes.

2.7 Proto-Japanese

43

(20) *ami-piki 'net-pull' > abiki [a"biigii] 'trawling' *yama-miti 'mountain-path' > yamadi [jamand}] 'mountain-path' *mura-nusi 'village-master' > murazi [muran3i] proper name *yama-ni-t:wi 'mountain-GEN-bird' > yamadori [jamandori]

2. 7.1.2 Distribution of OJ mediae and liquid The mediae and the liquid (/r/) did not occur word initially. There were, however, suffixes with initial media or liquid, for example, be- (verb extension~ necessitive), dani (restrictive particle~ 'at least~ even'), ga (case particle~ genitive), ra (noun suffix~ plural). For the mediae, this does not seem to reflect an original phonotactic restriction, but rather the secondary origin of the mediae. It is on the other hand a feature common with the so-called Altaic languages not to allow word initial liquids. SJ loanwords were readily taken into the language with initial media or liquid (7.2), and there were a few examples of them already in OJ (4.2.2)~ there is also a single example of a mimetic with initial/hi in OJ, bisibisi 'sniffling'.

2. 7.1. 3 PJ syllable final nasals For many OJ forms with a media it is not possible to recover a source with a vowel between nasal and tenuis. Some such forms may reflect contraction of a tenuis with a directly preceding syllable final nasal: (21)

*tunpu 'grain' > tubu *kanti 'rudder'> kadi *pinsa 'knee' > piza

It is not possible to recover internally the nature of these nasals. For the forms above, 1*-n/ is posited, but there is no reason to suppose that pre-OJ did not have /* -m/ or /* -IJI in addition to/* -nl. There is nothing to say that piza is not from *pimsa or *piiJsa. Recently it has been proposed that pJ in addition to /m, n/hadsyllable initial *IJ, reflected as OJ /0/ or in some cases /g/.

2. 7.1.4 PJ glides Table 2.6 shows one of two main views on the pre-history of the OJ syllable initial glides, /w, y/, viz. that they go back to pJ /*w, *y/. On this view pJ did not have voiced stops, but a competing hypothesis reconstructs pJ /*b, *d, *g, *z I which subsequently weakened and are reflected as OJ glides and zero: /w, y, 0, 0/. Positive evidence for this competing view is scarce.

44

2 OJ phonology

2. 7.2

Vowels and diphthongs

Until recently the main view held that there were only four primary pJ vowels: /*i, *a, *u, *~/reflected as OJ li, a, u, of. This view was adopted, for example, in Martin (1987), but recently more vowels are reconstructed. We here show the seven primary vowels reconstructed by Frellesvig and Whitman (2008b), adapting the account there of developments between pJ and OJ

2. 7.2.1 Arisaka sLaw; distribution of primary vowels Arisaka sLaw (named after Arisaka Hideyo) is a restriction on the shape of OJ root morphemes: /Col(= Co 2) did not occur in a root morpheme with /Ca, Cwo, Cui. That is to say, there were (almost) no simple words of the structure /CoCa, CoCwo, CoCu, CaCo, CwoCo, CuCo/. Arisaka's Law is summarized in (22) for the direct reflexes of the pJ vowels: (22)

Neutral -i (<*i, *e) -ye (<*e)

A -0

(<*i:,

*~)

B -u (< *u, *o-) -wo (< *-o) -a(<*a)

This has been taken to mean that Japanese earlier had some form of 'vowel harmony', but such a restriction on the shape of root morphemes is different from the kind of vowel harmony which applies to the concatenation of stem and affix in 'vowel harmony languages' such as for example Turkish. Structurally the primary pJ vowels were organized as in (23), with restrictions holding on the co-occurrence of central and back vowels within a root morpheme: (23)

Neutral *i *e

Central *i: *~

Back *u *o *a

2. 7.2.2 Secondary vowels and diphthongs: OJ 1-wi, -el OJ /-wi, -e/ are thought to be secondary and to reflect contraction of falling diphthongs /*Vy/ or /*Vi/: (24)

Pre-OJ *uy *i:y *~y

*ay

OJ

> > > >

WI WI

e e

45

2.7 Proto-Japanese

The most instances of OJ /e, wi/ are found in alternations which have come to be known as 'apophonic'. For example, a number of nouns have alternating shapes with variation in the final syllable: One variant, the free form (known as roshutsukei (21±1ID) 'exposed form'), occurs in word final position, while the other, the compound form (hifukukei (~5ID) 'covert form'), usually occurs in compounds or derived forms. Traditionally the covert form is thought to represent an older or more original shape of the word. Similar alternations are found among verbs, reflecting similar contractions, see (3.5.1). (25)

wi- u

Exposed form mwi 'body' kamwi 'spirit'

Covert form mu-kapari ('substitute') 'hostage' kamu-kaze ('wind') 'divine wind'

wi- o

kwi 'tree' yomwi 'Hades'

ko-dati ('tree-stand') 'grove' yomo-tu-kuni ('Hades-GENland') 'land of Hades'

e-o

se 'back' me 'bud'

so-muku 'turn' moyasi 'bud, sprouting forth'

e-a

me'eye'

ma-pye ('side, direction') 'front' saka-duki ('cup') 'sake cup'

sake 'sake'

In such cases the exposed form ends in /-e, -wi/, which are thought to reflect contraction of /*Vy/ or /*Vi/ as shown above in (24). The source of the /*-y/ in the diphthongs that were contracted has been a matter of debate, with proposals including the ancestor of the particle i (see 3. 7.1.3), but the apophonic nouns are now generally thought to go back to consonant final shapes, with the final consonant being weakened and reduced to *y. In some cases it is not possible to identify the consonant, hence *s::1C, *maC. (26)

'body' 'tree'

mwi kwi

<*muy < *kiy

'back' 'eye'

se me

< *s::Jy <*may

<*mum (cf. J\lfK mom) < *kir (cf. J\lfK kirih 'stump'; counter for trees) < *s::1C <*maC

The final consonant of such forms is thought to have been deleted in composition before another consonant, e.g. *mum+ kapari => mukapari. That is the origin of the covert form which more directly reflects the original root vowel:

46

2 OJ phonology

(27)

'body' 'eye'

< *mu-C. .. <*ma-C ...

mu-

ma-

<*mum-C .. . <*mar-C .. .

As shown in (25), OJ /o/ takes part in two apophonic alternations, /o- -wil and /o- e/. This is thought to reflect contraction of two different vowels with 1*-y/ (or /*-i/) and that is the reason for reconstructing two central vowels, pJ /*i:, *~/,which later, after the lexicalization of the contracted forms, merged, most likely as /*~/, which later backed to give OJ lol. Thus, OJ lol has two different pJ sources, /*i:/ and/*~/, and the only reason we recognize this is that the two sources gave different outcomes when contracted with /*y/, with the result that OJ /of takes part in two alternations: (28)

pJ

*~,

*i: > pre-OJ

*~

> OJ o

The two types of covert forms in OJ lol are derived as in (29): (29)

'tree' 'back'

koso-

< *ki:-C. .. < *s~-C. ..

< *ki:r-C. .. < *s~C-C ...

More generally, the source of the apophonic alternations may be set out as in (30): (30)

u- wi o- wi o-e a-e

*u> u *i: > *~ > *~ > 0 *a> a

0

wi < *uy wi < *i:y e < *~y e < *ay

In addition to apophonic alternations there are a few cases of fairly transparent lexical contractions which involve /*Vi/ sequences, shown in (31). Such cases were important in formulating the hypothesis that OJ /-e, -wil generally reflect contractions of diphthongs, but in fact there are very few solid examples of such lexical contractions. (31)

*ui > wi *i:i > wi

*ai > e

*waku-iratukwo 'young-honourable.male' > wakwiratukwo *i:pi:-isi 'big-stone' (cf. OJ opo- 'big')> opwisi proper name *bn~-iri 'palace-enter' (cf. OJ tono 'palace')> toneri 'palace servant' *taka-iti 'high-market'> taketi proper name

47

2.7 Proto-Japanese

Also here OJ /of is involved in two outcomes of contraction with /*i/: (32)

'big' 'palace'

opwisi < *ipi-isi toneri < *t::1n::1-iri

*ipi > opo*t:m::J > tono

2. 7.2.3 Mid vowel raising The pJ mid vowels *e and *o raised as in (33): (33)

*e > -ye in final position (partial MVR) i elsewhere (full MVR) *o > -wo in final position (partial MVR) u elsewhere (full MVR)

That is to say, in nonfinal position pJ *e, *o merged with pJ *i, *u as OJ Iii and /e/, respectively, but were preserved as distinct only in final position. It appears that there were dialectal differences in what constituted 'final position' for these sound changes, ranging between root final, morpheme final, and word final. Mid vowel raising seems to have followed the course of a gradual phonetic diphthongization, eventually giving a high vowel, except in final position where the diphthongal realization was phonemicized as a diphthong, see (34): (34)

*o > *o > *e > *e >

[wo] > [wo > u] > [ie] > [ie > i] >

/wo/ (partial MVR) lui (full MVR) /ye/ (partial MVR) Iii (full MVR)

A few examples which show both outcomes are given in (35). The different developments of *yo(-)ri reflect different interpretations of an internal boundary. (35)

*me 'woman' > *e 'placenta' > *yo(-)ri 'behind'> *moko 'bridegroom'>

OJ mye- womina (< *wo-me-na 'small-fern ale-person') OJ ye- iro- 'of same mother' (< 'placenta-GEN') OJ ywo{-)ri- yuri 'from; behind' OJ mwokwo - mukwo > EMJ muko

Note that the effect of partial MVR. was neutralized in OJ for *e after coronals and glides, for *o after labials, and eventually in EMJ after all consonants, because of the merger of /Ce/ and /Cye/, and of /Co/ and /Cwo/ (see 7.3.2.1).

48

2 OJ phonology

(36)

pJ

pre-OJ

*Ce>

*Cye>

a.

b. *Co>

OJ Cye> I C = {p, b, m, k, g}

EMJ Ce

Ce> I C = {t, d, s, z, n, r, w (y)}

Ce

Cwo> I C = {t, d, k, g, s, z, n, r, y}

Co

Co> I C = {p, b, (m) w}

Co

*Cwo>

In addition to reflecting directly pJ !*of and /*e/, some instances of OJ 1-wo/ and /-ye/ are secondary in the sense that they derive from contraction of sequences of vowels, e.g. (37):

(37)

*ia > ye

*saki-ari 'bloom-is' > sakyeri 'is blooming'

*i;l > ye

*pi-;)ki 'sun-put'> pyeki proper name

*U;)

>

WO

*ua > wo

situ-ori 'native. weaving-weave' - sitwori 'type of cloth' *kazu-ap- 'number-join'> kazwope- 'count'

However, there are very few good examples of this, apart from a small number of fairly transparent grammatical morphemes involving contraction of /*-i-a-/:

(38)

-yer- stative < *-i-ar- 'INF-exist' -kyeku simple past nominal< *-ki-aku 'sPST-nominalizer' -kyeku adjectival copula nominal< *-ki-aku 'ACOP.ADNnominalizer' -kyem- simple past conjectural< *-ki-am- 'sPST-CONJ' -kyer- modal past< *-ki-ar- 'sPST-be' (or 'come'-exist) kyer- 'be wearing' (lexicalized stative of ki- 'wear') < *ki-ar- 'wear-exist' kyes- wear, respectful < *ki-as- 'wear-RESP' myes- see, respectful< *mi-as- 'see-RESP' kyesa this morning< *ki-asa 'this-morning'

49

2.7 Proto-Japanese

2. 7.2.4 Lexical distribution of /Cwo, Cye, Ce, and Cwil Reflecting their pre-history, the syllables /Cwo, Cye, Ce, Cwi/ are restricted in the OJ lexicon. This is shown by the following result of a count of the text occurrence of different syllable types in the Man'yosha (Ohno 1980: 151ff.), here maintaining the orthographic three-way distinction between ko, otsu, and neutral syllables. (39)

Ci1 3,160

Ci 6,103

Ci 2 370

Ce1 686

Ce 2,299

Ce2 853

Co1 1,030

Co 3,631

Co2 5,280

Ca 12,120

Cu 6,415

A frequency count in running text gives no real picture of lexical distribution, but these figures do indicate that /Cwil (= Ci2), /Cye/ (= Ce 1), ICe!, and /Cwo/ (= Co 1) were relatively infrequent compared to /Ci/ (= Ci1), /Cal, /Col(= Co 2), and /Cui. In terms of distribution in the lexicon (as opposed to running text), /Cwi/ is infrequent and is almost exclusively found in morpheme final position. /Cye/ and /Ce/ are also rather infrequent. By far the most occurrences are in morpheme final position, but while simple words with nonfinal /Cye/ and /Ce/ are rare, they are not exceptional. /Cwo/ is not lexically infrequent, making up approximately 25 per cent of occurrences of Co 1 and Co2 in the words listed in Omodaka et al. 1967, but most occurrences of /Cwo/ are in morpheme final position (see Martin 1987: 60-2 for a list of words with /Cwo/). These distributional facts reflect (a) that all OJ /Cwi, Ce/ are etymologically secondary, and (b) that /*o/ and /*e/ merged with /u/ and li/ in nonfinal position and only gave /wo/ and /ye/ in root, morpheme, or word final position.

2. 7.2.5 Changes between proto-Japanese and Old Japanese An approximate relative chronology of the changes which took place between pJ and OJ may be outlined as in ( 40) overleaf:

50

2 OJ phonology

(40) a.

Contraction of falling diphthongs Pre-OJ > OJ *ui WI *ii WI *:li e *ai e

b. Merger of central vowels pJ {*i, *:l} > *:l c.

Mid vowel raising pJ *e > OJ i, -ye

d. Contraction of rising diphthongs pre-OJ *ia > *ya > OJ -ye pre-OJ *i:l > OJ -ye e.

Mid vowel raising pJ *o > OJ u, -wo

f.

Backing of central vowel pre-OJ *:l > OJ o

g. Contraction of rising diphthongs pre-OJ *ua > *wa > OJ -wo > OJ -wo pre-OJ *uo REFERENCES General: Mabuchi 1971, Martin 1987, Wenck 1959. Greek vowels: Horrocks 1997: 102ff. Phonetic reconstruction based on Chinese sound values for ongana: Lange 1973, Miyake 2003a. Ko-otsu distinctions: Lange 1973, Miyake 2003a, Unger 2007; Co1 vs Co2: Hayata 1998, Matsumoto 1984. Consonants: Frellesvig 1995: 58-86, Martin 1987: 1-36. Prenasalization and medial voicing: Wenck 1959. /s, z/: Kobayashi 1981, Ogura 1998, Sandness 1986. Proto-Japanese: Frellesvig and Whitman 2008a, Whitman 1985.

3

Grammar

3.1

Verbs

In a segmentational model, Japanese verb forms may be described in terms of the following structure, with five main morphemic layers:

2 4 3 5 root- derivative -auxiliary verb- auxiliary- flective

(1)

(2)-( 4) are examples of OJ verb forms: (2) a.

tuku '(it) attaches'

b.

(3) a.

tuk-

5 u

attach-

CONCL

tukeme 'can/will attach it!'

b.

(4) a. b.

4

5

tuk-

2 e-

m-

e

attach

TRANSITIVIZER

CONJ

EXCL

maywopikinikyeri 'had become frayed!' 3

4

4

maywopi-

ki-

ni-

kyer-

fray

come

PERF

MPST

5 CONCL

The central inflectional morphology is realized by flectives, expressing obligatory inflectional categories for which any verb inflects (3.1.3), and by auxiliaries expressing optional categories for which verbs can inflect (3.1.4). As seen in (4), more than one auxiliary can be attached to a verb, in which case the auxiliaries appear in the ordergivenin Table3.2 and(21) below (3.1.4.1 and 3.1.4.5). Of the four layers, or positions, in (1) after the root, only flective has obligatory expression, reflecting that any instantiation of a verb appears in an inflected form. 51

52

3 OJ grammar

3.1.1

Derivatives

A number of pre-OJ verbal derivatives may be reconstructed, most of which were, however, fully lexicalized and unproductive by the time of OJ. The verbal derivatives shown in (5) may be assumed to have been fairly productive in OJ; they mostly specify or switch transitivity, but also include an iterative/ continuative formant. In addition there are a number of derivatives which form verbs from other parts of speech, e.g. -bwi- and -sabwi- mentioned in 3.4.1.2 below. (5)

-(a)s- transitive; -(a)r- intransitive; -e- opposite transitivity (cf. 3.5.1); -(a)p- activity verb.

3.1.2

Auxiliary verbs

As mentioned, root + derivative constitute the lexical base. More than one lexical base can combine to form a compound. Lexical composition must, however, be distinguished from affixation of auxiliary verbs, which to some extent correspond to, or at least overlap with, what have been termed 'syntactic compounds' (e.g. Kageyama 1999). Grammaticalized auxiliary verbs attach to the infinitive of a main verb, specifying amongst other things various motional, directional and aspectual meanings, but including also expression of social deixis. They are set off from auxiliaries (3.1.4) by generally being used as free lexical verbs in addition to their grammaticalized use as auxiliary verbs. The degree of grammaticalization varies and it is therefore difficult in some cases to distinguish between verbs commonly used in lexical compounds and fully grammaticalized auxiliary verbs. Among the fully grammaticalized auxiliary verbs are those in (6): (6)

-ko- [ < 'come'] 'come to ... ', -yuk- [ < 'go'] 'continuative, go on .. .ing; gradually become ... more and more', -ide[< 'emerge'] 'VERB out'; -ap- [<'meet'] 'reciprocal'; -e[< 'get'] 'be able to'

-tamap- [<'receive'] 'respect', 1 -mas-[< imas- 'exist, respect'] 'respect'; -mawos- [<'say to/tell the emperor/a superior; ask for permission (of a superior)'] 'humble', -matur- [<'offer to spirit or emperor'] 'humble' 1

The auxiliary verb -tamap- follows rather than precedes the respect auxiliary -(a)s- (cf. 3.1.4.3), e.g. oka-si-tamapite (MYS 5 .813) 'put-HON-HON'. This is an exception to the sequence set out in (1) above. The number of attested examples is very small, but this indicates that -tamapshould perhaps be interpreted as an auxiliary rather than an auxiliary verb; also in EMJ, -tamap- follows supporting respect auxiliaries.

3.1 Verbs

3.1.3

53

Inflected verb forms: obligatory categories

In the segmentational model given above in (1), inflectional categories are represented by 5, flective. Table 3.1 lists the inflected forms for each of the eight OJ conjugational verb classes, noting the verb base (basic stem) at the top of each column. These are the paradigmatically opposed forms for which any lexical verb inflects (except that some verbs will not form an imperative). See 3.4 about the conjugation classes. The first thing to note is that inflection in OJ is almost exclusively for zyntactic, modal and conjunctional categories (as opposed to NJ which also inflects, for example, for tense and aspect). The primary distinction is between finite forms, which can conclude a main clause (3.1.3.1), and non-finite forms, which conclude nonfinal clauses (3.1.3.2)~ there is also a nominal form (3.1.3.3). The following summarizes main functions of each category and gives for reference the forms of the verb sin- 'dies'.

3.1.3.1 Finite verb forms Conclusive (sinu) This form has two functions: (a) The conclusive is used for concluding declarative main clauses, see (7b). This is the function from which its name derives. (b) It is also used before most extensions (3.6), final particles (3.7.5), and with the concessive conjunctional particle to(mo) (3.7.4). It should be noted, however, that it is quite rare to find simple verbs (without auxiliaries attached) in the conclusive form concluding a main clause within the OJ corpus~ most occurrences of simple conclusive verb forms are found with extensions or particles. The conclusive form is unspecified for- neutral with regard to- tense, aspect, or mood~ thus all auxiliaries, including those expressing tense, aspect, and mood, have a conclusive form. Labels such as 'indicative' or 'non-past' which are sometimes used about this form in OJ (and EMJ and early LMJ) are therefore misleading. Adnominal (sinuru) This form had several uses: First, the function after which the form is named is to modify a noun, that is to say, as the verb in an adnominal clause: 'who dies~ dying'. In this function it contrasts with the conclusive form:

(7) a.

b.

tama-no-ura ni asari suru tadu Tama-no-ura DAT forage.INF do.ADN crane 'A crane fishing in the Tama-no-ura' (MYS 15.3598) nwosnna ga saki ni ipori su, ware pa Noshima GEN cape DAT hut do.CONCL, I TOP 'Me, I make a hut on the cape ofNoshima' (MYS 15.3606)

54

3 OJ grammar

Table 3.1 Injlectedfonnsfor the eight OJ verb classes

Base Finite Conclusive Adnominal Exclamatory Imperative Neg. conject. Optative Prohibitive

QD

r-irr

n-irr

kak-

ar-

sin-

'paint, write'

'exist, be'

'die'

kaku kaku

ari aru are are arazi arana na ari so

sinu sinuru sinure sine sinazi sin ana nasini so

kakite kakitutu kakaba kakeba kakedo

ari arite aritutu araba areba aredo

sini sinite sinitutu sinaba sinureba sinuredo

kakaku

araku

sinuraku

kake kakye kakazi kakana na kaJci so

Non-finite Infinitive Gerund Continuative Conditional Provisional Concessive

kaJci

Nominal

LB

UB

UM

k-irr

s-irr

ake-

okwi-

mi-

ko-

se-

'open'

'rise'

'see'

'come'

'do'

Finite Conclusive Adnominal Exclamatory Imperative Neg. conject. Optative Prohibitive

aku akuru akure ake(yo) akezi akena naake so

oku okuru okure okwi(yo) okwizi okwina naokwi so

(mi) miru mire mi(yo) mizi mina nami so

ku kuru kure ko kozi kona

su suru sure se(yo) sezi sena na se so

Non-finite Infinitive Gerund Continuative Conditional Provisional Concessive

ake akete aketutu akeba akureba akuredo

okwi okwite okwitutu okwiba okureba okuredo

mi mite mitutu miba mireba miredo

lei kite kitutu koba kureba kuredo

si site situtu seba sureba suredo

Nominal

akuraku

okuraku

miraku

kuraku

suraku

Base

3.1 Verbs

55

Second, the adnominal also functions to form headless nominalizations, meaning both (i) 'the one who dies'; and (ii) 'that someone dies'. In OJ, the former use is almost only found in pseudo-cleft constructions, e.g. (8). The use of the adnominal in complement clauses ('that ... ') is very rare in OJ, where generally the nominal form was used, see 3.1.3.3. However, in EMJ both of these nominalizing uses of the adnominal became frequent. (8)

[kadi no oto suru] pa ama-wotomye kamo oar GEN sound do.ADN TOP fisher-girl Q 'the ones making the oar-sounds, is that the fisher -girls?' (MYS 15.3641)

Third, the adnominal form is also used as the predicate in exclamative or interrogative main clauses, often but not always in correlation with a focus or interrogative particle, so, namo, ya, ka (see 3. 7.2 and in particular 8.9). Finally, it is used as the verb of a subordinate clause concluded by some conjunctional particles, e.g. ni, wo, and those derived from nouns (see 3. 7.4). In both these last two functions, the adnominal may be regarded as being nominalized and it has also been proposed (Shibatani, p.c.) that the use of this form in adnominal clauses is a nominalizing use, which would make all uses of this form nominalizing. The functions of the OJ conclusive and adnominal largely remained unchanged in EMJ, but underwent important changes in LMJ, see 12.6.1. Exclamatory (sinure) This form is mainly used to form the predicate of an exclamative main clause, sometimes on its own, but often in correlation with the focus particle koso (see 8.9.2). The exclamatory can also function as the predicate in a subordinate clause, often translated with one of a num her of conjunctional meanings: 'if, when, although, because'. Usually it is accompanied by some modal or other particle. It also functions as a combining stem, selected by some suffixes, see 3.4.4. Imperative (sine) This is used as the direct imperative and to form directive clauses when followed by the purposive conjunctional particle to ('in order that ... '). Negative Conjectural (sinazi) The negative conjectural is semantically the negative counterpart of the conjectural auxiliary -(a)m- and has the same range of meanings. It generally means 'I don't want to, I shan't die; you shouldn't, mustn't die; he probably, surely won't die'. It is also often used to form directive clauses when followed by the purposive conjunctional particle to ('lest ... '). Traditionally, the negative conjectural has, due to its semantic affinity with the conjectural auxiliary -(a)m- (3.1.4.1 0), been regarded as an auxiliary, -(a)zi, with the sole form -zi functioning as conclusive, adnominal and exclamatory (reflecting the traditionally recognized types of sentence conclusion, neutral and in correlation with a focus or interrogative particle).

56

3 OJ grammar

However, nothing supports that view: the negative conjectural is simply an inflected word form. Optative (sinana) OJ has three sets of optative forms, see (10), of which only one is listed in Table 3.1. They express the wish of the speaker, predominantly (but not exclusively) about actions of a 1st, 2nd, or 3rd person, respectively. With the exception of -(a)namu, which is used throughout EMJ as a general optative, these forms disappear from the language in the transition from OJ to EMJ; -(a)namo is thought to be the older variant and the source (through mid vowel raising, see 2. 7.2.3) of -(a)namu which was more frequent already in OJ. (9)

sin ana sinane(mo), sinani(mo) sinanamo, sinanamu

'I would like to die' 'I wish you'd die/ please die' 'I wish he'd die/ let him die'

Prohibitive (na-sini-so) The prohibitive expresses a negative command 'don't .. !'. There are three variants, see (1 0), of which only na ... so survived into EMJ. The prohibitive may well be thought to be a construction rather than an inflected word form, and in EMJ longer stretches than a single verb can occur between na and so, but this does not seem to have been possible in OJ. Including the prohibitive as a word form allows us to posit a circumfixal formant, the only one of its kind in pre-modem Japanese, na-. . .-so, which surrounds the infinitive used as stem. (With se- 'do', the base rather than the infinitive is used: na-se-so 'don't do!'.) Etymologically, na is probably a negative adverb and it may be thought that the use of na as a prohibitive prefix is the older construction which later came to be supported by so (cf. also 3.1.5); so is thought to reflect *s:), the historical root, used as imperative, of the verb se- 'do'. This finds some support in the long form na-. . .-sane which seems to involve an older fossilized optative form, sane 'I wish you'd ... ', of the ancestor of OJ se-, again involving the older root. Note also the prohibitive final particle na (3.7.5). (10)

na-sini-so na-sini-sone na-sini

'don't die!' '(please) don't die!' 'don't die!'

3.1.3.2 Non-finite verb forms All stages of Japanese, including OJ, are characterized by the existence of a number of non-finite verb forms which are differentiated by the type of conjunction that holds between the clause they conclude and a following clause. Infinitive (sini) The non-finite verb forms include one coordinate form, the infinitive, whose main use is to form a nonfinal predicate coordinate with a

3.1 Verbs

57

following predicate. A number of examples point to an earlier stage where the infinitive was subordinate, 2 a function which at the OJ stage had largely been taken over by the gerund (which is newer in the language). Derived from and segmentally identical with the infinitive is the substantive noun, e.g. itupari 'lie, deceit' (itupar- 'to lie'); in EMJ and in later periods, the infinitive and its derived noun are attested with different prosodic shapes (see 7.4.4.3) and this is also thought to have been the case for OJ. The infinitive is also used as a combining stem, both in composition and selected by some suffixes (see 3.4.4.1). The remaining non-finite verb fonns are subordinate. Gerund (sinite) The gerund is a neutral subordinate verb fonn, unspecified for the type of conjunction that holds between its own and a following higher clause. The use of the word 'gerund' for such subordinate adverbial verb fonns is now customary in the description of many languages, although it is somewhat misleading to those more familiar with the use of 'gerund' to designate a nominalized verb fonn in for example Latin grammar. For Japanese this nomenclature is found first in the work of the Portuguese Jesuits around 1600 (e.g. Rodrigues's Arte da lingoa de Iapam, see 10.2.2.2), based, presumably, on the similarity in meaning with the Portuguese gerilndio, an adverbial fonn which historically derives from the ablative of the Latin gerund. Rodrigues, however, also classifies the fonn as 'participle'. The other subordinate, non-finite fonns are specified for some type of conjunction with the higher clause. Continuative (sinitutu) The continuative expresses continuation, repetition, contemporaneity, or by extension, concession; many of its meanings can be captured by English 'while'. Conditional (sinaba) The conditional concludes a conditional subordinate clause 'if'. Provisional (sinureba) The provisional concludes a provisional, temporal, or causal subordinate clause, most of which can be rendered by English 'as'. Concessive (sinuredo) The concessive concludes a concessive subordinate clause 'although, even though'; the concessive is often followed by the particle mo with no discernible difference in meaning. 3.!.3.3 Nominal This form expresses nominalization: 'the fact of dying; that someone dies'. It was a very frequent and prominent fonn in OJ (also of adjectives, see 3.2.2.2) 2

For example (i) which is found in MYS 2.85, 5.867, K 88:

(i)

kimi ga yuki ke naga-ku my.lord GEN go.JNF day long-ACOP.INF 'many days have passed since you, my lord, left'

nari-nu become-PERF.CONCL

58

3 OJ grammar

whose basic function was abstract nominalization ('the fact that ... '), (11 a), often used in complement clauses ('that .. .'), (llb-c), but also for example as a nominalized exclamative predicate, (11 d). See 3.4.4.4 about the formation of the nominal form. ( 11) a.

aki tukeba momiti tiraku autumn arrive.PROV autumn.leaves scatter.NMNL 'The scattering of the autumn leaves when autumn comes' (MYS 19.4161)

b.

[nara no asuka wo miraku] si yo-si mo Nara COP.ADN Asuka ACC see.NMNL EMPH good-ACOP.CONCL EfOP 'It is good to see Asuka which is in Nara' (MYS 6.992)

c.

[wa ga kokoda sinwopaku] sira-ni GEN this.much yearn.NMNL know-NEG.INF I 'Not knowing that I would yearn this much' (MYS 19.4195)

d.

kimi ga mi-koto wo motile kaywopaku lord GEN HON-word ACC hold.GER arrive.NMNL 'It [the branch of jewel-pine] arrives bearing the words of you, my lord!' (MYS 2.113)

The nominal form disappears as a productive form in the transition to HMJ where it only survived in set phrases with a small number of verbs (cf. 9.1. 7). From EMJ the adnominal form takes over the function of use in complement clauses, but in OJ there was clear functional differentiation between the nominal and the adnominal form, the latter not being used in abstract nominalizations and only very rarely in complement clauses, of which (12) is an example. Wrona (2008) details the uses of the nominal and adnominal forms in OJ, and more generally describes the OJ system of complementation, including the use of the formal noun koto 'thing, fact' both in abstract nominalization and in complement clauses. (12)

ari-kyeru] sira-ni [yama-mori no mountain-guard GEN be-MPST.ADN know-NEG.INF 'not knowing that there was a mountain guard' (MYS 3.401)

3.1. 4

Auxiliaries: optional categories

Auxiliaries are inflecting suffixes. Most auxiliaries belong to and inflect according to one of the eight verbal conjugation classes, but a few have idiosyncratic conjugations; none has the full range of inflected forms exhibited by lexical verbs, see below (3.1.4.2).

3.1 Verbs

59

Whereas simple inflection (3.1.3) mainly expresses modal and conjunctional categories, auxiliaries express central morphological categories such as voice, tense, aspect, and mood. Thus the two sets of categories complement each other. However, an important difference is that inflection is obligatory, whereas the expression of the categories of the auxiliaries is optional. That is to say, any OJ verb form expresses one and only one of the set of paradigmatically opposed inflected forms in Table 3.1, but no auxiliary need be present. This means that a simple verb form, without an auxiliary, is unmarked(unspecified) with regard to the categories expressed by auxiliaries. For example, a simple verb form is unmarked with regard to tense and can well refer to deictic past time. Further, more than one auxiliary can be attached to a verb, in which case they occur in the order given in Table 3.2 (from the top down; cf. also (16) in 3.1.4.5). This affords the categories expressed by auxiliaries a different position within the morphological system from those expressed, obligatorily, by simple inflection. The combination of auxiliaries is subject to systematic restrictions, reflecting paradigmatic subsystems which are described in 3.1.4.5.

3.1.4.1 Formation on lexical verbs Table 3.2 shows the OJ auxiliaries attached to verbs from the eight verb classes. As the auxiliaries themselves inflect, the resulting forms are shown in the basic stem, not in a word form. 3 As may be seen, there are some systematic restrictions on combinations between lexical verb classes and auxiliaries. The most important is that the bigrade verbs (which are diachronically secondary) do not combine with the respect, passive or stative auxiliaries; see Frellesvig (2008). Other non-formations concern small classes with few members and may not be significant. It is noteworthy, however, that n-irr verbs combine with no aspectual auxiliaries, and that no UM verb combines with the -(i)nperfective. The latter is probably related to the fact that most (original) UM verbs are quite transitive; cf. 3.4.3.1. It is not surprising that r-irr verbs, which are all semantically stative, do not combine with the stative auxiliary. 3.1.4.2 1n.flectedforms The auxiliaries have the inflected forms shown in Table 3.3, with the auxiliaries ordered in four subgroups: respect, voice (causative and passive), aspect/ negation (perfective, stative, negative) and tense/mood (modal past, simple past, conjectural, subjunctive). It is significant that the auxiliaries do not have 3

Note that the two perfective auxiliaries generally combine with different verbs (see 3.1.4.6.2); hence we have added sale- 'bloom', tuke- 'attach', andyodi- 'uproot' to the QD, LB, and UB columns. The only UM verb attested with the stative auxiliary is ki- 'put on', so kyer- 'be wearing' is added to the UM column. k-irr and s-irr verbs have limited and partly irregular formations with the simple past, see (3.4.3.2).

60

3 OJ grammar Table 3.2 OJ auxiliaries aitached to verbs from the eight verb classes

QD kak-

r-irr ar-

n-irr sin-

sinasimesinayesinare-

Respect

kakas-

Voice Causative Passive Passive

kakasimekakayekakare-

arasimearayearare-

AspecUNegation Perfective Perfective Stative Negative Negative

kakite(sakin-) kakyerkakankakazu

aritearinaranarazu

sinansinazu

Tense/Mood Modal past Simple past Conjectural Subjunctive

kakikyerkakiki kakamkakamasi

arikyerariki aramaramasi

sinikyersiniki sinamsinamasi

LB ake-

UB okwi-

Respect Voice Causative Passive Passive AspecUNegation Perfective Perfective Stative Negative Negative Tense/Mood Modal past Simple past Conjectural Subjunctive

UM mi-

k-ilT ko-

myes-

s-irr seses-

akesime-

okwisime-

misimemiye-

kosime-

sesime-

(tukete-) aken-

(yodite-) okwin-

mite-

akenakezu

okwinokwizu

(/eyer-) minmizu

kitekin/eyerkonkozu

sitesinsersensezu

akekyerakeki akemakemasi

okwikyerokwiki okwimokwimasi

mikyermiki mimmimasi

lei/eyer(kosi) komkomasi

sikersiki semsemasi

3.1 Verbs

61

Table 3.3 Injlectedfonns of OJ auxiliaries Respect -(a)s-

Conclusive Adnominal Exclamatory Imperative Optative Prohibitive

su su se se sane na VERB-si so

Infinitive Gerund Conditional Provisional Concessive

si site saba seba sedo

Nominal

saku

Passive Causative -(a)sime-

-(a)ye-

-(a)re-

Conclusive Adnominal Imperative Negative conjectural

simu simuru sime(yo)

yu yuru

ru ruru

yezi

rezi

Infinitive Gerund Continuative Conditional

sime simete simetutu

ye yete yetutu yeba

re rete retutu reba

yuraku

ruraku

Nominal Perfective

Stative

Negative

-(i)n-

-(i)te-

-yer-

-(a)zu- -(a)n-

Conclusive Adnominal Exclamatory Imperative Optative

nu nuru nure

tu turu ture teyo lena

yeri yeru yere yere

zu nu ne

Infinitive Gerund Continuative Conditional Provisional Concessive

ninile nitutu naba nureba nuredo

te-

yeri

zu-ni zute- nito

teba tureba turedo

yeraha yereba yeredo

zupa. neba nedo

Nominal

nuraku

turaku

yeraku

naku

nana

62

3 OJ grammar Table 3.3 (cont.) Modal past -(i)kyer-

Simple past -(i)ki

Conjectural -(q)m-

Subjunctive -(q)masi

Conclusive Adnominal Exclamatory

kyeri kyeru kyere

ki si sika

mu mu me

masi masi

Conditional Provisional Concessive

/eye reba /eye redo

seba- kyeba sikaba sikado

medo

Nominal

kyeraku

siku- kyeku

maku

maseba

the full range of inflected forms exhibited by verbs. They also differ a great deal among themselves with regard to the categories they inflect for, but note that the differences are greater between the four subgroups than within them. 3.1.4.3 Respect The respect auxiliary belongs to the QD conjugation. It regularly combines with QD and s-irr (and according to reading tradition also r-irr verbs). There is a small number of lexicalized forms, (13a) with UM verbs, (13b) derived from roots underlying bigrade verbs, and finally (13c) with slightly irregular formation. (13) a.

myes- < *mi-as-; cf. mi- 'see' (also in the compound -(a)si-myes'RESP-RESP')

kyes- < *ki-as-; cf. ki- 'put on' (rare, only in the form kyes-eru (RESP-STAT.ADN 'which you are wearing') and in mi-kyesi (deverbal noun; 'honourable garment')) b. koyas- < *koyV-as-; cf. ko(y)i- 'lie (down)' (UB) nas- < *na-as-; cf. ne- 'sleep, lie' (LB) c.

omopos- <= omop- 'think (of), believe' (reading tradition also gives the regular form omopas-) kikos- <= kik- 'hear, listen' (the regular form kikas- is also attested) oros- <= or- 'weave' siros- <=sir- 'rule; know' (cf. siro 'castle') (there is only one OJ attestation of siros-, in the form sirosi-myes- (EN 8); sirosgained in usage in EMJ. The regular OJ forms are siras-, sirasimyes-.)

3.1 Verbs

63

The respect auxiliary is used to express 'subject exaltation', cf. 12. 7.1, which however also has basic deictic functions, as illustrated in (14): (14)

ko no woka ni na tuma-su kwo this GEN hill DAT greens pick-RESP.ADN child 1pye kikana nora-sane home ask.OPT tell-RESP.OPT '(You,) girl who picks greens on this hillside, I want to ask your home. Please tell me!' (MYS 1.1).

3.1.4.4 Voice; causative and passive The causative and passive auxiliaries belong to the LB conjugation. The OJ causative is used to express causation and respect (see 12.7.1.2). It is formed on verbs from all conjugation classes. In EMJ, aLB causative -sase- was used alongside, and eventually replaced, -(a)sime- (see 8.4.1). Early formations involving a precursor of the new emerging causative are found already in OJ, in particular in the following forms:

(15)

kise- <= ki- 'wear'; mise- <= mi- 'see'; apase- <= ap- 'meet'; kikase- <= kik- 'hear'; sirase- <=sir- 'know; rule'; opose- <= op- 'carry'.

The OJ passive functioned as a pure passive, a medium voice, and as a potential. OJ had two competing passive markers of which -(a)ye- was used rather more frequently than -(a)re- (their roles are reversed in EMJ into which -(a)ye- only survived in a number of lexicalized forms (kikoye-, miye-, omopoye-) and in reading glosses to Chinese texts, see 9.1.6). -(a)ye- was used with QD, n-irr, r-irr and UM verbs; -(a)re- with QD, n-irr, and r-irr verbs. Passives were not formed on verbs from other conjugation classes. -(a)ye- thus had a wider use than -(a)re-, both in terms of distribution and in terms of frequency. The straightforward interpretation of these facts is that -(a)ye- was the earlier passive marker and was replaced by innovative -(a)re-. There are a few lexicalized passive forms with a slightly irregular formation: kikoye- <= kik- 'hear'; omopoye- <= omop- 'think' (found alongside omopaye-, but was much more frequent). The passive auxiliary variant -rare-, attaching to vowel base verbs, does not appear in the language until EMJ (see 8.4.1) and forms no part of OJ, although it is included in some grammars. Furthermore, an OJ passive auxiliary *-rayeis traditionally posited, said to attach to UB and LB and sometimes also to s-irr and k-irr verbs. This is based solely on four occurrences of the set phrase i no nerayenu (i 'sleep', no genitive) in MYS 15 (3665, 3678, 3680, 3684), whose nerayenu is traditionally analysed as ne-raye-nu sleep-PASS-NEG 'not

3 OJ grammar

64

be able to sleep'. However, the non-formation of passives on bigrade verbs is entirely systematic (see 3.5.1). The singular form neraye- must be explained otherwise, e.g. as the passive of a verb ner-, 4 or as an idiosyncratic analogical formation~ it cannot alone form the basis for positing a general variant *-rayeas part of the grammar of OJ. Note that the causative and the passive were paradigmatically opposed in OJ (could not co-occur in a verb syntagm), as opposed to later stages of the language where they combined in some of their functions (see 8.4.1).

3.1.4.5 Aspect tense, negation, andmood The core aspectual and temporal morphology of OJ is expressed by the five auxiliaries -(i)n-, -(i)te-, -yer-, -(i)ki, -(i)kyer- (usually an auxiliary -(i)tar- is also posited in OJ grammars, but see 3.1.4.7.3). There is a remarkable lack of agreement about the basic function of each of these auxiliaries, which form part of the core verbal morphology of OJ, and about their mutual relations, probably related to the focus of most pre-modern Japanese grammars on a positive identification of the meaning of individual grammatical morphemes, instead of seeing them as parts of a (number of) system(s). The morphological system of tense and aspect is inextricably interwoven with that of negation and mood. The full system also includes the negative -(a)zu - -(a)n-, the conjectural -(a)m-, and the subjunctive -(a)masi. The system exhibits the following paradigmatic relations within three categories. Thus, only one member from each category can occur in a verb syntagm. If more than one category is present, they occur in the order given here. (16) a.

b.

Aspect!Negation

Perfective -(i)te- ~ -(i)nStative -yerNegative -(a)zu- -(a)n-

Tense/Mood (i)

Modal past -(i)kyerSimple past -(i}ki Subjunctive -(a)masi

(ii)

Modal past -(i)kyerConjectural -(a)mSubjunctive -(a)masi

As indicated by the occurrence of the subjunctive and the modal past in two categories each, these three categories combine to form two subsystems which 4

Ner- would in turn probably be best understood as a form, e.g. a lexicalized stative, of the root, *nV-, which diachronically underlies LB ne-. This root is also reflected in the lexicalized respectful verb nas- 'deign to sleep', and possibly also in the noun i 'sleep' ?< *ni ( cf. Whitman 1985: 34 about palatalization and loss of /*d, *z, *n, *r/ before /*if).

3.1 Verbs

65

may be illustrated as follows~ the horizontal dimension shows combination, the vertical dimension shows selection (mutual exclusion): (17)

Aspect/Negation Stative

I I

Perfective

Negative

Tense/Mood Modal past

1\ \I

Simple past

Conjectural

Subjunctive

3.1.4.5.1 Aspect and negation Aspect and negation together form one subsystem with the perfective as the pivot: the perfective (-(i)te- - -(i)n-) forms an opposition with and never combines with the stative (-yer-). The perfective is also opposed to and never combines with the negative (-(a)zu-(a)n-). Finally, the negative and the stative do not combine, but this seems, at least diachronically, to be a secondary opposition (cf. 3.1.4.6.1). See further 3.1.4.6 about the perfective, 3.1.4.7 about the stative and 3.1.4.8 about the negative. 3.1.4.5.2 Tense and mood Tense and mood form a more complex second subsystem: the modal past (-(i)kyer-) is opposed to and never combines with the simple past (-(i}ki) or the conjectural (-(a)m-). Likewise, the subjunctive (-(a)masi) is opposed to and never combines with the simple past or the conjectural. The simple past does, however, combine with the conjectural -(a)m- in the fairly autonomous combined shape -(i)kyem-. See 3.1.4.9 for the functions of the past tenses. Note that the modal past and the conjectural, as subjective modals (c£ 3.1.4.9, 3.1.4.10), do not form a conditional, as opposed to the simple past and the subjunctive~ see Table 3.3(d) in 3.1.4.2. 3.1.4.5.3 Combination As implied by the illustration of the system in (17), aspect/negation combines relatively freely with tense/mood, as in the following examples:

3 OJ grammar

66 (18)

'hear' 'bloom' 'meet' 'stand' 'be satisfied' 'be satisfied'

kiki-te-kye-mu saki-ni-kyeri ap-yeri-ki tat-era-masi aka-ni-kye-mu aka-zu-kyeri

Aspect/Negation PERF PERF STAT STAT NEG NEG

Tense/Mood SPST-CONJ MPST SPST SUBJ SPST-CONJ MPST

Note, however, that examples of the negative with the past tense auxiliaries are very rare and that the negative combines with the conjectural and subjunctive only in its analytic forms formed with the existential verb ar- (see 3.1.4.8.1), and that it thus strictly speaking does not directly combine with those auxiliaries.

3.1.4. 6 Perfective

3.1.4.6.1 Functions The perfective combines two main functions: Aspectually, a perfective is a verb form that views a situation in its entirety, including beginning, middle, and end. This whole can be looked at from both ends, as it were, so as with perfectives in other languages both (a) ingressive (inceptive) and (b) completive uses are possible, as in (19). This is the origin of the opposition with the stative. (19) a. b.

naki-nu 'begins to sing', naki-tu 'begins to sing' tiri-nu 'has fallen', mi-tu 'has seen'

The other main function of the perfective is to assert or affirm the state of affairs expressed by the verb. This is the origin of the opposition with the negative (and with negation in general: the perfectives do not form a negative conjectural). In this function the perfective often combines with the conjectural, -te-rn-, -na-m-, e.g. (20a), or assumes a modal form, e.g. (20b) with the perfective in the optative. Diachronically this seems to be the primary function of -(i)te-- -(i)n-, reflecting the fact that they derive from the ancestors of the OJ copulas to - ni, see (3.5.2). (20) a.

miti-ki-na-mu rise-come-PERF -CONJ '(the tide) will surely rise' (MYS 2.121)

b.

kari-tena cut-PERF.OPT 'I want to cut (seaweed)!' (MYS 2.121)

3.1 Verbs

67

3.1.4.6.2 Distribution of the variants -(i)te- and -(i)n- The distribution of the two perfective auxiliaries has traditionally been described as being based on the transitivity of the host verb: (21)

-(i)te- is used overwhelmingly with transitive verbs -(i)n- is used mainly with intransitives

There are a fair number of exceptions to this generalization, however, and for that reason some scholars hold that -(i)te- and -(i)n- (already) in OJ were distinct morphemes expressing separate grammatical categories (for example Sandness (1999: 54ff.): -{i)n- 'punctuality', -{i)te- 'recent past'). However, a number of distributional facts make it clear that -(i)te- and -(i)n- do belong closely together in OJ; -{i)te- and -{i)n-: (22) a. b. c. d.

are mutually exclusive; occupy the same position in a verb syntagm; do not co-occur with the stative and with the negative; exhibit mostly the same inflected forms.

On the other hand, the precise details of the differences in use between -(i)te- and -(i)n- remain unclear. There are clearly discernible, strong tendencies in their distribution in terms of sem anti co-syntactic properties of the host verb, particularly when refining this in terms of split intransitivity: 5 (23)

Transitives -{i)te-

Intransitives unergatives unaccusatives -{i)te-{i)n-

There are, however, also exceptions to this generalization, and in particular there are some verbs which occur with both perfective auxiliaries. First, the following grammatical verbs are found with both perfective auxiliaries: light verbs: se- 'do' and ar- 'exist'; auxiliary verbs: -ko- 'come to ... ', -yuk- 'continuative'. However, in such cases, the choice of perfective auxiliary does not depend on the grammatical verb, but rather on the lexical host verb. More importantly, there is a small number of lexical verbs which are attested with both suffixes:

5

Though it appeared for the first time, unattributed, in print in Frellesvig (2001: 14), this suggestion is due to John Whitman (p.c.).

68

3 OJ grammar

(24)

ip- 'say, tell, call', kakure- 'hide (oneself\ die (respect)', ko'come', mi- 'see, look', miye- 'see.PASS: be visible/seen, seem; come.RESP', nak- 'give voice, let out a cry, cry', ne- 'lie down, sleep/lie with, sleep', ori- 'go down, descend', pum- 'step (on), stamp (on); walk', sinwop- 'praise, long for',yuk- 'go'.

It should first of all not be overlooked that there may have been variation between OJ speakers in the use of the perfectives and in the definition of criteria for distribution. More importantly, however, as pointed out by Sorace (2000), auxiliary selection based on split intransitivity is not exclusively a matter of invariable lexical properties of host verbs, but proceeds along a hierarchical scale and includes semantic properties deriving from the clauses/predications in which the perfective occurs. 6 It is single argument (intransitive) verbs which have no or a low lexical specification of telicity, affectedness, agentivity, and volitionality which exhibit variation in the choice of perfective auxiliary. In particular, agentivity and volitionality seem to be relevant for OJ. 3.1.4. 7 Stative

3.1.4.7.1 Function The stative presents a situation as a state. This includes the result of an action and the stative is thus in some respects similar to a perfect. The Japanese stative has been known by many different names, e.g. resultative, durative, progressive, perfect, or imperfect, but the term 'stative' adopted here captures better the basic meaning. A stative is not formed on r-irr verbs, which are inherently stative; in OJ the two lexical n-irr verbs, sin- and in-, did not form a stative. OJ had a morphological stative auxiliary and a periphrastic stative construction. 3.1.4.7.2 Morphological stative auxiliary: -yer- The stative auxiliary -yerbelongs to the r-irr conjugation. It has a systematically restricted distribution: It is regularly formed on QD, s-irr and k-irr verbs, and it is also found with the UM verb ki- 'put on' (kyer- 'be wearing, have on'), but is not formed on other verb classes, in particular not the lower and upper bigrade classes, reflecting that they are derived and secondary and most likely younger in the language (cf. 3.5.1). Etymologically, the morphological stative results from univerbation of an analytic construction consisting of the infinitive followed by the existential verb ar-, with contraction of the two contiguous vowels (see 2.7.2.2): *saki-ar- 'bloom-be'> sakyer- 'be in bloom, be blooming'. 6

Washio (2002) is the first to point out explicitly that selection of perfect auxiliaries in Germanic and Romance involves similar issues to the selection of perfective auxiliaries in OJ.

3.1 Verbs

69

3.1.4. 7.3 Periphrastic stative: -(i)te ar-, -(i)tar- Whereas the morphological stative was not used with all verb classes, a periphrastic stative construction was available for that purpose, also formed with an existential verb, but following the gerund of the verb, e.g. panarete ar- 'be separate' (<= panare'growlbecome separate'). This construction is synonymous with the morphological stative. It is likely that the periphrastic stative construction arose in complementation of -yer- to be used with the secondary (bigrade) verb classes as they emerged and later was generalized to occur with all classes. The periphrastic stative was also formed on the continuative+ ar-: VERB-tutu ar-, and it was also formed using one of the exalted synonyms of ar-, for example mas-, imas-. The pattern of forming statives by gerund+ existential verb is pervasive through the history of Japanese and it is still the way statives are formed in cNJ where the new existential verb i- is used (see 12.4). In the later part of OJ we find phonological fusion of -te ar-, e.g. nokorite ar- 'remain'=> nokoritar- (<= nokor- 'be left behind'). The auxiliary -(i)tar- is a prominent and important auxiliary in MJ~ through EMJ it increased dramatically in frequency at the expense of -yer- which it eventually replaced in LMJ, and it is the ancestor of the NJ past tense flective -(i)ta. However, although -(i)tar- is included in most grammars of OJ, there is little positive basis for positing it as an independent grammatical form: OJ !-tar-! is a simple phonological shortening of the periphrastic stative VERB-tear-=> VERB-tar-, derived by synchronic phonological rules (see 2.6.1, 3.4.2.1.2). 3.1.4. 7.4 Analytic progressive Activity verbs formed an analytic progressive with the verb wor- 'be sitting' used directly after the infinitive: tomosi wor- 'be burning (something), be lighting a fire'~ wor- is the opaque lexicalized stative form of wi- 'sit down', see further 3.4.2.1.

3.1.4.8 Negative The negative exhibits a suppletive paradigm in OJ, with variation between two basic variants, -(a)n- and -(a)zu-, even within some forms, namely infinitive and gerund, column (a) in Table 3.4. Infinitive -ni and gerund -nito are rare already in OJ and both forms went out of use in the transition from OJ to EMJ, to give the suppletive paradigm in column (b) in Table 3.4 (which also includes a new gerund variant). The OJ paradigm reflects a transitional stage, with innovative -(a)zu based variants being in the process of replacing, or having replaced, some of the forms in an earlier QD type paradigm. See 3.1.4.8.3 about the reformation of the inflectional paradigm of the negative. Morphophonologically, the negative paradigm is interesting. Its two basic variants belong to entirely different conjugational types: -(a)n- conjugates like a QD verb (with a notable irregular gerund). There are no lexical QD

70

3 OJ grammar Table 3.4 OJ transitional paradigm and EMJ suppletive paradigm of the negative

(a) OJ transitional paradigm

EMJ suppletive paradigm

Conclusive Adnominal Exclamatory

zu nu ne

zu nu ne

Infinitive Gerund Concessive Provisional Conditional Nominal

ni -zu nito - zute (- zuni - zusite) nedo neba zupa

zu zute- de nedo neba zupa

naku

naku

(b)

-n- bases and the negative auxiliary -(a)n- is thus the only QD type -n base in the language. On the other hand, the infinitive and conclusive -zu is in several respects like -ku, the infinitive of the adjectival copula (cf. 3.2.2.4.1). Both serve as a stem in the formation of further forms, cf. gerund -zute and -kute, and conditional -zupa and -kupa (note also the variant gerund -zuni and the extended gerund -zusite). Finally, -(a)zu is syntactically like the existential verb ar- in using the infinitive in conclusive function.

3.1.4.8.1 Analytic forms Also like -ku, negative -zu may be extended with ar-, especially to combine with auxiliaries which never or only rarely attached directly to the negative auxiliary. For example, the negative combines with the conjectural-(a)m- and the subjunctive -(a)masi- only in the ar- extended form: -zu ara-m-, -zu ara-masi, and although direct combinations of the negative auxiliary with tense auxiliaries are attested (see 3.1.4.5.3), they are very rare and -zu ari-kyer- or -zu ari-ki- are more frequent. The analytic forms gave full morphophonological versatility to the negative auxiliary and use of analytic forms continued through LMJ. In some cases, -zu ar- phonologically fused to give -zar-, e.g. -zara-m-, -zara-masi, -zariki (see further 3.4.2.1.1 and 3.4.2.1.2). In the extended forms, negation could apply recursively: miyezaranaku ni <= mi-ye-zu ara-naku ni see-PASS-NEG exist-NEG.NMNL COP.INF 'although it is not (the case) that (she) is not visible'.

3.1.4.8.2 Negative rhetorical questions In negative rhetorical questions the negative could occur at the end of a verb syntagm, rather than in its usual position (preceding tense auxiliaries), taking wide sentential scope. Thus, there are a few examples with the modal past -kyerazu ya, e.g. (25). In this use, the

3.1 Verbs

71

Table 3.5 Reformation of the OJ paradigm ofthe negative (a) OJ QD -(q)nConclusive Adnominal Exclamatory Infinitive Gerund Concessive Provisional Conditional Nominal

(b) OJ innovating fonns (nu) nu ne ni

zu <

*ni-su

zu <

nito ( *nite) nedo neba

zute <

*ni-su *ni-su-te

*naba

zupa. <

naku

*ni-su-pa

(c)

EMJ zu nu ne zu zute -de nedo(mo) neba zupa. naku

negative could even combine with the perfective, in which case it was attached through extension with ar-, e.g. (26): (25)

kadura ni su be-ku hair. decoration do.CONCL COP.INF NEC-ACOP .INF nari-ni-kyera-zu ya become-PERF-MPST-NEG.CONCL Q 'shouldn't it have been made into a hair-decoration?' (MYS 5.817)

(26)

saku be-ku nari-nite ara-zu ya bloom.CONCL NEC-ACOP.INF become-PERF.GER exist-NEG.CONCL Q 'shouldn't it have started to bloom?' (MYS 5.829)

3.1.4.8.3 Reformation of the paradigm of the negative As mentioned above, -(a)zu was in OJ close to completing the replacement of some of the variants in-(a)n- within the negative paradigm, cf. the EMJ paradigm (repeated here in column (c) of Table 3.5). As -(a)n- conjugates like a QD verb (with a notable irregular gerund) we may hypothetically posit an earlier regular QD conditional and a regular gerund, marked with'*', and we include a conclusive -nu in brackets (paralleling a regular QD verb), see Table 3.5, column (a). The -(a)zu forms involve a formant *su attached to the the QD type infinitive -ni, as shown in (33b). This *su is usually identified with the conclusive su of the verb 'to do', but that is doubtful given the syntactic and morphological peculiarities of *su which, as mentioned above, behaves much like the adjectival -ku; see further 3.5.2 about this.

72

3 OJ grammar

In order to understand the motivation for the reformation of the forms of the negative it is revealing to compare the forms of the negative of QD, LB and UB verbs with the corresponding forms of the -n- perfective, see Table 3.6. We see that with QD verbs, the inflected forms of the perfective and negative are distinct throughout. However, with the bigrade verbs some of the inflected forms of the perfective are homophonous with forms of the -(a)nnegative~ those forms are underlined. It is precisely those forms which were replaced by -zu based variants. Thus, the paradigm of the negative may be thought to have been reformed in order to avoid homonymy between the forms of the negative and the perfective in the bigrade conjugations. Recall (3.1.4.5.1) that the negative and the perfective took part in a paradigmatic opposition: it is not tolerable to have widespread homonymy between the forms of two members of a grammatical opposition. The reformation of the negative paradigm thus seems to have been motivated by the emergence and lexicalization of the bigrade conjugations (see 3.5.1).

3.1.4. 9 Simple and modal past Of the two OJ past tense auxiliaries, modal past -(i)kyer- belongs to the r-irr conjugation whereas simple past -{i)ki has its own idiosyncratic paradigm, with both s-and k- initial forms (see Table 3.7~ see further 3.5.2). In the conclusive, conditional and nominal form there are variant forms of the simple past, of which -kyeku and -kyeba are very rare~ the few examples are limited to the early poetry in the Kojiki and the Nihon shoki and there are no examples in the Man 'yoshii. On the other hand, -siku and -seba are frequent and survive into EMJ. The simple past combines with the conjectural -(a)m-: -kye-m-, whereas the modal past does not~ nor does it form a conditional. This difference highlights the basically modal function of the modal past.

3.1.4.9.1 Functions We will here briefly discuss the function(s) of the two auxiliaries -(i)ki and -(i)kyer- and the differences between them. Sometimes the term 'retrospective' is preferred instead of 'past', but an important function of both -{i)ki and -(i)kyer- is to refer to deictic past time. The modal past is more complex in usage than the simple past. It is widely acknowledged to involve some modal element, hence the name given it here, but this has proved difficult to pinpoint. Often a general difference is said to be one of directly experienced (-(i)ki) versus indirectly experienced, or evidential, past (-(i)kyer-). More specifically, the difference is traditionally set out as in (27). The simple past is usually said to refer to something the speaker has experienced himself, but on the other hand it can also be used in historical accounts. The modal past has a variety of uses: it is said to be used about hearsay, or about sudden realization, and also with some more general emphatic or exclamatory force, in which case it frequently has no past reference. Another

3.1 Verbs

73

Table 3.6 OJ -n- perfective and negative forms of QD, LB and UB verbs QD sak- 'bloom' Perfective -(i)n-

Negative -(a)n-

Negative -(a)zu

Conclusive Adnominal Exclamatory Infinitive

sakinu sakinuru sakinure sakini

sakanu sakanu sakane sakani

sakazu

Gerund Concessive Provisional Conditional

sakinile sakinuredo sakinureba sakinaba

sakanito sakanedo sakaneba

sakazute

*sakanaba

sakazupa

Nominal

sakinuraku

sakanaku

Perfective -(i)n-

Negative -(a)n-

Negative -(a)zu

Conclusive Adnominal Exclamatory Infinitive

tokenu tokenuru tokenure tokeni

tokenu tokenu tokene tokeni

tokezu

Gerund Concessive Provisional Conditional

tokenite tokenuredo tokenureba tokenaba

tokenito tokenedo tokeneba

tokezute

*tokenaba

tokezupa

Nominal

tokenuraku

tokenaku

Perfective -(i)n-

Negative -(a)n-

Negative -(a)zu

Conclusive Adnominal Exclamatory Infinitive

okwinu okwinuru okwinure okwini

okwinu okwinu okwine okwini

okwizu

Gerund Concessive Provisional Conditional

okwinite okwinuredo okwinureba okwinaba

okwinito okwinedo okwineba

okwizute

*okwinaba

okwizupa

Nominal

okwinuraku

okwinaku

sakazu

LB toke- 'melt (intr.)'

tokezu

UB okwi- 'arise'

okwizu

74

3 OJ grammar Table 3.7 OJ modal past and simple past auxiliaries Simple past Modal past -(i)kyer-

-(i)ki

EMJ

Conclusive Adnominal Exclamatory

kyeri kyeru kyere

ki si sika

ki si sika

Conditional Provisional Concessive Nominal

kyereba kyeredo kyeraku

seba- kyeba sikaba sikado siku- kyeku

seba sikaba sikado (siku)

kye-m-

kyem-

Conjectural

set of uses, however, is said to have no modal meaning, but rather to resemble a perfect 'has/had been . . .'. 7 (27)

Simple past -(i)ki

Modal past -(i)kyer-

direct experience, historical accounts

hearsay, sudden realization, emphatic or exclamatory force; perfect

3_1.4.9.2 'Perfect' Examples where -{i)kyer- is said to function as a perfect on closer inspection turn out to involve an entirely different formation, namely the combination of the auxiliary verb -ko- 'come to', cf. (4) with the stative auxiliary -yer- giving the form -{i)kyer- 'has come to', which is homophonous with the modal past. This is for example clearly illustrated by (28):

7

On the basis of the traditional description outlined in (27), the Turkish 'direct experience' versus 'indirect experience' past tenses are often cited as a close parallel to the OJ simple past versus modal past (as in Shinzato 1991), see (i). Although this parallelism is widely accepted (e.g. Takeuchi 1999: 101ft'.), it turns out on closer inspection to be quite general and to consist in some modal distinction among past tenses. There is no close fit between the uses of the two OJ past tenses and the two Turkish past tenses. (i)

OJ

-(i)ki

-(i)/eyer-

Turkish

-DI Direct experience

-mi., Indirect experience

3.1 Verbs

(28)

75

kamwiyo god.age

ywori ipi-tute-kuraku ABL say-transmit-come.NMNL sworamitu yamato no kuni soanng Yamato GEN land sumyekamwi no itukusiki kuni ruling.deity GEN august.ADN land no sakipapu kuni kotodama word.spirit GEN bless.ADN land katari- tugi ipi-tugapi-kyeri tell-continue.INF say-continue-come.STAT.CONCL

pa

TOP

to COMP

'It has been recounted down through time since the age of the gods: that this land of Yamato is a land of imperial deities' stern majesty, a land blessed by the spirit of words.' (Levy 1981) (MYS 5.894)

Viewed in its entirety, (28) involves the framing construction where reported speech is not only followed by a complementizer and a verb of reporting, but also introduced by the same, or a similar, verb in the nominal form, schematically shown in (29) with the verb ip- 'say':

(29)

... ipaku say.NMNL

'reported speech'

to ipCOMP say

In (28), the verb introducing and concluding the reported speech unambiguously incorporates the auxiliary verb -ko- 'come to': -kuraku is the nominal form of the auxiliary verb -ko-, and not of the modal past -(i)kyer-, whose nominal form is -kyeraku (cf. Tables 3.1 and 3.3d). The form concluding the reported speech therefore also involves -ko-, here in combination with the stative auxiliary. Leaving aside the question of whether the modal past is etymologically related to ko- 'come', the auxiliary verb -ko- and the auxiliary -(i)kyer- were clearly different at the OJ stage, both in their relative positions within a verb syntagm, cf. maywopi-ki-ni-kyeri ((4) above) which contains both -ko- and -(i)kyer-, but on different sides of perfective -(i)n-, and in their inflected forms, cf. Tables 3.1 and 3.3d. But the combination of the auxiliary verb -ko- with the stative auxiliary -yer- is homophonous with the modal past auxiliary -(i)kyer-. It turns out, then, that what have until now been understood as different uses or functions of one grammatical morpheme, -{i)kyer-, in fact are two entirely different formations which are homophonous. One is the auxiliary

76

3 OJ grammar

verb -ko- combined with the stative auxiliary -yer-~ the other is the modal past auxiliary, whose functions will be described immediately below. The main environment where a distinction between the two can be formally made is in combination with a perfective auxiliary: (-ko- + -yer-) => -kyerdoes not further combine with a perfective auxiliary (as stative and perfective are paradigmatically opposed and do not combine), whereas the modal past does combine with both of the perfective variants: -(i)te-kyer- and -(i)ni-kyer- unambiguously involve the modal past. Clearly, more detailed research is called for in order to chart the differences between these two different formations in other environments and their further history within Japanese.

3.1.4.9.3 Speaker commitment There is little or no positive evidence for anything 'indirect' or 'evidential' about the OJ modal past -(i)kyer-. In particular, there are no good examples of the modal past used about hearsay in OJ: in examples where a hearsay interpretation is not simply imposed by translators any sense of reporting actually resides in a lexical verb of reporting and not in the auxiliary -(i)kyer-. The element which is common to the uses of the modal past (captured by 'sudden realization', 'emphasis' in the traditional description), may best be understood as speaker commitment, imparting a sense of 'I tell you', as in examples such as (30) and (31), which clearly have nothing to do with indirect experience. Such a modal use of a past tense is well known from other languages, e.g. Norwegian, (32). In both (31) and (32), the past tense component is semantically overridden by the modal component, but that does not make the Norwegian past tense, or the OJ modal past, less of a past tense. It simply means that they have modal uses which pragmatically can dominate.

(30)

wegusi ni ware wepi-ni-kyeri smiling.sake DAT I get.drunk-PERF-MPST.CONCL 'I became drunk on the sake of smiles!' (KK 49)

(31)

miyabwiwo ni ware pa elegant.fellow COP.INF I TOP 'what an elegant fellow I am!' (MYS 2.127)

(32)

sa.finjeg var na.l so fine I was now 'how nice I look!'

ari-kyeri exist-MPST. CONCL

3.1 Verbs

77

By contrast, the simple past -(i)ki is a neutral, simple preterite which places a situation before the time of speaking, or some other temporal reference point. 8 Positing explicitly an element of subjectivity as the defining characteristic of the modal past is nothing new. Takeuchi (1999: 10 Iff.) shows that that interpretation goes back to the eighteenth century. Two facts which underpin this interpretative intuition are, first, that the modal past does not, as opposed to the simple past, have a conditional form, cf. Table 3.3d above. This is a morphological property the modal past shares with the conjectural auxiliary -(a)m-; both are subjective modals, and truly subjective modals do not form conditionals. Second, the difference between the two past tenses is clearly brought out by a striking difference in their distribution in the two sets of OJ prose texts: (33)

Simple Past Modal Past

Senmyi5 addresser focused: from an emperor/empress

Norito addressee focused: to the spirits

+ +

+

The modal past does not occur in the Norito at all, while it is used in the Senmyo. In narrative terms, the main difference between these two sets of texts is one of address: the Norito are addressee focused, first of all being addresses to the deities/spirits; Senmyo on the other hand are addresser focused, being addresses from an emperor or empress. There is thus less scope for subjectivity and speaker commitment in the Norito which focus on the addressee. In the Senmyo, on the other hand, a clear pattern of usage may be observed which utilizes the distinction: the modal past is used in direct or indirect speech with some element of speaker commitment, e.g. (34)-(35). Conversely, the simple past is used in frame descriptions, almost entirely about divine or imperial past action (and with a respectful verb form), e.g. (36): (34)

sikaredomo una pa akiraka ni Nakamaro ga however now TOP clear COP.INF Nakamaro GEN itupari ni ari-kyeri sirite ... to deceit exist-MPST. CONCL know.GER COP.INF COMP 'However, knowing clearly now that it was Nakamaro's deceit'

(SM28) • The distinction between the simple and modal past in OJ is similar to the modal distinction proposed - if not wholly uncontroversially - by Durst- Andersen (1996) for the two past tenses in Italian, with the passato prossimo resembling the OJ modal past in expressing speaker involvement, as opposed to the passato remoto which is said simply to report neutrally, like the OJ simple past.

78

3 OJ grammar

(35)

imasite arne sikaredomo sumyera to no however emperor COP.INF exist.RESP.GER heaven GEN sita no maturigoto wo kikosi-myesu bottom GEN ruling ACC perform-RESP.ADN koto pa itapasi-ki ikasi-ki koto ni thing TOP laborious-ACOP.ADN hard-ACOP.ADN thing COP.INF ari-kyeri exist-MPST. CONCL 'However, ruling the land as emperor was laborious and hard!' (SM23)

(36)

takama no para yu amori-masi-si sumyera high.heaven GEN plain ABL descend-RESP-SPST.ADN emperor 'an emperor who descended from the high heavens' (SM 2)

The Senmyo are recited texts with several embedded layers of narration and they might thus be said to be all direct speech, but the point is that the simple past is used in what is presented as frame description, or asserted to be part of common knowledge, i.e. shared and generally accepted truth, whereas the modal past expresses speaker involvement and commitment. Thus, the distinction between the two OJ past tenses has nothing to do with directness of experience; instead they differ in the expression of speaker commitment.

3.1.4.10 Conjectural and subjunctive The conjectural belongs to the QD conjugation. The subjunctive has its own conjugation, the few forms of which resemble the simple past. The conjectural is the least specific of the modal auxiliaries and is very frequent in the OJ texts. Its uses fall into two overall categories: (a) conjecture: probability and necessity; (b) volition: intention and wish. Note that the combination of the simple past and the conjectural -kye-m- is used only to express conjecture 'probably was, probably did', not volition. Like modals in many languages, the conjectural can also be used to refer to future time with little or no modal meaning; the conjectural is therefore sometimes termed a 'future tense'. Followed by the conjunctional particles to or ni it is used to form purposive adjunct and complement clauses 'that, so that'. The subjunctive expresses a counterfactual proposition, most often contingent on a counterfactual condition in the frame VERB 1-(SUBJ}COND, VERB 2SUBJ 'if it were the case that VERBh then it would be the case that VERB 2 '.

3.2

3.1.5

Adjectives

79

Verbal prefixes

Like later stages of Japanese, OJ used prefixes with both nouns and verbs. Nominal prefixes include honorific mi-, which is reflected in later stages of the language (see 12.7.1.1), or endearing mutu- which was no longer productive at the OJ stage, but only found in lexicalized forms such as mutu-tama 'friendly spirit'. Verbal prefixes include what are usually glossed as 'emphatic' prefixes, e.g. uti- which seems to derive from the infinitive of the lexical verb ut- 'strike'. Use of such deverbal emphatic prefixes is prevalent also in later stages of the language. Quite different from such prefixes which are found throughout the history of Japanese, we find in OJ what seem to be remnant uses of grammatical prefixes expressing functions which from OJ onwards were regularly expressed by suffixes: (a) reciprocal api- (api-makura-mak- 'pillow-spread (= sleep) together') < ap- 'meet'; cf. the derivational suffix -(a)p- (3.1.1) and the auxiliary verb -ap- (3.1.2). (b) Stative ari- (ari-mat- 'be waiting, keep waiting'); < ar- 'exist'; cf. the morphological stative -yer- < *-i ar- (3.1.4.7). (c) Potential e- (e-yuk- 'can go');< e- 'get'; cf. both the use of e as a freer adverb introducing negative potentiality and the potential auxiliary verb -e- (3.1.2). (d) Prohibitive na- has no transparent lexical source and was in OJ usually used in combination with the suffix -so (see 3.1.3.1); cf. also the prohibitive final particle na (3.7.5). The use of these prefixes which were lost in later stages of the language suggests that prefixation perhaps was used more widely in pre-OJ to express grammatical functions. This is not typical of verb final languages and could be taken to suggest that a change from preverbal to postverbal grammatical affixation could have been part of a wider change in syntactic typology towards verb final syntax. Less transparent, but perhaps part of the same phenomenon, are two verbal prefixes, i- and sa-. These two prefixes were frequent in OJ, but later disappeared from the language. Their function is not agreed upon (and they are often said to be 'emphatic'), but it has recently been suggested that i- and sa- are related to the expression of the transitivity or agentivity of the verbs they prefix to, so that i- attaches to active verbs (i-yuk- 'go') and sa- to inactive verbs (sa-ne- 'sleep'), see Yanagida and Whitman (2009); Russell points out that i- is attested mainly with verbs of movement (2006: 141 ff.).

3.2

Adjectives

Adjectives are nominal roots or stems. The stem was usually followed by an auxiliary (see 3.2.1), but it could also be used with a fair amount of independence. The bare stem could be used exclamatorily, usually reinforced by an

80

3 OJ grammar

interjection or an interjectional particle, e.g. (37). This use is still found in NJ, often with an elongated final vowel, takaa 'that's expensive!' (37) a. ana omosirwo 'how wonderful!' (Kogoshili) b. oso ya, kono kimi 'he is stupid, this man!' (MYS 9.1741) The stem was also used attributively and adverbially, first by compounding, (38a), (39a); second by juxtaposition, (38b), (39b); third with a form of a copula, adnominal no, tu in (38c), or infinitive ni in (39c); there are no examples of the stem of a shiku adjective adnominalized with tu. (38b) also exemplifies the adjective stem forming a nexus with a preceding noun, either as a compound pa-biro 'wide-leaved', or as a phrase yo no topo 'of old age'. In such cases, the resulting structure was used to modify a following noun. (38) a. yasu-i 'good, sound sleep' pasi-duma 'dear, beloved wife' b.

fyo no topo] pi to 'person of old age; [age GEN distant] person' (NSK62) [mi ga posi] kuni 'the country I long to see; [see GEN want] country' (KK 58) pa-biro kumakasi 'wide-leaved great oak' (KK 91)

c.

topo no kuni 'distant country' (MYS 15.3688) ikasi no miyo 'abundant reign' (EN 27) topo tu pito 'distant person' (MYS 17.3947)

(39) a.

tika-duku 'approach; close/near-touch/stick to'

b. paya ko 'come quickly!' (MYS 15.3636) c. 3.2.1

ko ni kaki 'paint thickly' (KK 42) Inflectionalforms; a4fectival copula

However, adjectives were usually predicated, adnominalized, or adverbialized by means of particular formants which attached to the adjective stem and inflected for many, but not all of the categories of the verbal inflection, see Table 3.8. These formants were suppletive and composite in their inflection. Because of their function (to adverbialize, adnominalize and predicate adjectives) and shared morphological development with the copula (12.2.3) we interpret these formants as forms of a restricted copula ('adjectival copula') and gloss examples accordingly, e.g. samu-ku cold-ACOP.INF 'coldly'. This

81

3.2 Adjectives Table 3.8 OJ adjectival copula forms

Finite Conclusive Adnominal Exclamatory-! Exclamatory-2

si

lei sa kyere

Non-finite Infinitive-! Infinitive-2 Gerund-! Gerund-2 Conditional-! Conditional-2 Provisional-! Provisional-2 Concessive-! Concessive-2

ku mi kute milo kyeba kupa kyeba kyereba kyedo kyeredo

Nominal

kyeku kye-naku kye-m-

Negative nominal Conjectural

interpretation of the formants attaching to the adjective stem finds further support in its relations to some other grammatical forms (see 3.5.2). There are two slightly different classes of adjective. Adjectives with stem final-si use a zero allomorph of the conclusive formant -si, i.e. kwopwisi and not *kwopwisi-si, see (50) which shows the combination of the conclusive, adnominal and infinitive forms of the adjectival copula with these two classes of adjective. Traditionally the two classes are termed' ku adjectives' and' shiku adjectives' in reference to the infinitive, topoku versus kwopwisiku, although in fact the si of the shiku adjectives is part of the stem and the infinitive ending in both cases is -ku. Shiku adjectives involve an adjectival derivational formant -si, see 3.2.4. A small subclass of shiku adjectives had stems in -zi rather than -si, see 3.2.4.2.

Ku adjective

(40) Conclusive Adnominal Infinitive-}

topo- 'far, distant'

Shiku adjective kwopwisi- 'dear, beloved'

topo-si topo-ki topo-ku

kwopwisi kwopwisi-ki kwopwisi-ku

82

3 OJ

grammar

In OJ the syntactic specialization which we find in EMJ between the conclusive, -si, and adnominal, -ki, forms of the adjectival copula was not yet firmly established (see 3.2. 2.1). In addition many categories exhibit variation between two competing forms. In some cases, this variation seems to reflect that one set offorms, the innovative forms in (b) in (41), was replacing another, those in (a); compare with the EMJ forms in column (c). (41) Exclamatory Conditional Provisional Concessive

3.2.2

a. OJ older sa kyeba kyeba kyedo

b. OJ innovative kyere kupa kyereba kyeredo

c. EMJ kere kupa kereba keredo

Core forms

The following forms of the adjectival copula were used widely and may be said to constitute the core forms in OJ: (42)

Conclusive Adnominal Exclamatory Infinitiv e-1 Infinitive-2 Nominal

si ki sa ku mi kyeku

The remaining forms were not frequent: -kyere is attested once, but in EMJ this form became frequent. -kute was very rare in OJ (only about a dozen wholly or partly phonographically attested examples in the Man 'yoshil), but common in EMJ. The conditional, provisional and concessive forms are infrequent in the texts, but the innovative set became widely used in EMJ. There are about a dozen attestations altogether of -kyeba and -kyereba; there is a single example of -kyeredo and a handful of -kyedo. The adjectival copula did not generally combine directly with auxiliaries. The only exception was the conjectural, -(a)m-: -kye-m-. There are approximately twenty or so attestations in the Man 'yoshil, comprising different inflected forms of -(a)m- (conclusive, adnominal -kye-mu, exclamatory -/eyeme, nominal-kye-maku). In addition, there are a few examples of the nominal form of the negative -(a)naku: -kye-naku, indicating that the adjectival copula was earlier used more freely with negative -(a)n-: *-kye-n-. These forms disappeared altogether in the transition to EMJ and were instead formed analytically, see 3.2.3 below.

3.2 Adjectives

83

3.2.2.1 Conclusive and adnominal The conclusive and adnominal are generally associated with those two functions, (43)-(44). Like the verbal adnominal, the adjectival copula adnominal was also used as a nominalized form in headless nominalizations, (49). However, the syntactic specialization between conclusive, -si, and adnominal, -ki, of the adjectival copula was not complete by OJ and there was some overlap in usage. There are many examples of a shiku adjective in what appears to be the conclusive form modifying a noun, e.g. (45), but the traditional interpretation of such examples is as a compound of adjective stem and noun. There is, however, a small num her of examples with ku-adjectives where a conclusive form of the adjectival copula is used to modify nouns, e.g. ara-si wo 'tough man' (ara- 'wild, violent, tough') or fully lexicalized yo-si nwo 'good moor; Yoshino (place name)' (yo- 'good'). ( 46) is an illustrative pair of examples with kagurwo- 'black (of hair)' (< ka 'hair', kurwo- 'black'). Examples such as these demonstrate that what can unambiguously be identified as a conclusive form of the adjectival copula could be used to modify a noun. On the other hand, such examples are rare in the OJ texts. As in later stages of the language, and as is the case with verbs, the adnominal form may be used predicatively, but in those cases it either correlates with a focus particle as in later stages of the language, (47) (note that koso in OJ correlates with the adnominal of the adjectival copula, not the exclamatory; see 8.9.2), or is associated with interrogative or exclamative function, ( 48) (cf. 8.9.4). (43)

a ga mune ita-si I GEN heart painful-ACOP.CONCL 'my heart aches' (MYS 15.3767)

(44)

kurwo-ki mi-kyesi black.ACOP.ADN HON-clothes 'black clothes' (KK 4)

(45)

topotoposi kosi no kuni far.far.away Koshi GEN land 'the far far away land ofKoshi' (KK 2)

(46)

kagurwo-ki kami 'black hair; black-ACOP.ADN hair' (MYS 5.804) kagurwo-si kami 'black hair; black-ACOP.CONCL hair' (MYS 16.3791)

(47) a.

aki to ipeba kokoro so ita-ki autumn COMP say.PROV heart FOC painful-ACOP.ADN 'when talk is about the autumn, my heart aches!' (MYS 20.4307)

84

3 OJ grammar b. nwo wo piro-mi kusa koso sige-ki moor ACC wide.INF grass KOSO abundant-ACOP.ADN 'with the moor being wide, it is the grass which is abundant' (MYS 17.4011)

(48)

koromode no puru toki mo wa ga GEN dry.ADN time ETOP I GEN sleeve na-ki not.exist-ACOP.ADN 'there is no (not enough) time for my sleeves to dry!' (MYS 10.1994)

(49)

[ipu subye no taduki mo na-ki] say.ADN way GEN means ETOP not.exist-ACOP.ADN a ga mwi nari-kyeri body COP-MPST.CONCL I GEN 'The one with no means of expressing himself was me!' (MYS 18.4078)

pa TOP

3.2.2.2 Nominal The use of the nominal form does not differ much from the use of the verbal nominal form, (50)-(51). Like it, it was an important and frequent form in OJ, but disappeared in the transition to EMJ. (50)

yononaka no u-kyeku tura-kyeku this.world GEN sad-ACOP.NMNL hard-ACOP.NMNL 'the sadness and hardness of this world' (MYS 5.897)

(51)

kwopwisi-kyeku ke naga-ki mono-wo long.for-ACOP.NMNL day long-ACOP.ADN although 'although it has been a long time that I have longed for you' (MYS 17.3957)

3.2.2.3 Exclamatory -Sa, which in later stages of the language functioned as a nominalizer, was in OJ a predicative, exclamatory form, usually occurring in the frame: NOUN gal no ADJ-sa or, more rarely, VERBIADJ ga ADJ-sa (52a, b). As opposed to the verbal exclamatory (and the rare exclamatory-2, -kyere), -sa could not be used as a stem for forming other inflected forms ( cf. 3.4.4.3). (52) a.

wakayu young.trout

turu fish.ADN

imwora girls

WO

ACC

mi see(BASE)

ramu PCONJ.ADN

3.2 Adjectives

85

pito

no tomosi-sa person GEN enviable-ACOP.EXCL '(how) enviable are those who will see the girls fishing young trout!' (MYS 5.863)

b.

ware

yuwe

ni

I

reason

COP.INF worry.CONCL

omopiwabu

ga

imo

ramu PCONJ.ADN

kanasi-sa

my.beloved GEN dear-ACOP.EXCL '(how) dear is my beloved who will be worrying for my sake!' (MYS 15.3727) c.

punabito wo miru ga tomosi-sa row.ADN boatsman ACC see.ADN GEN enviable-ACOP.EXCL '(how) enviable it is to see the rowing boatsman!' (MYS 15.3658)

kogu

d. apu be-ki yosi no na-ki meet.CONCL NEC-ACOP.ADN means GEN not.exist-ACOP.ADN ga sabusi-sa GEN lonely-ACOP.EXCL '(how) lonely it is that there is no means of meeting her' (MYS 15.3734)

3.2.2.4 The infinitives 3.2.2.4.1 Infinitive-1 The infinitive has two main functions: (a) adverbial, modifying a verbal, (53)-(55); (b) non-finite (56)-(58). In (57), taputwo-ku is the predicate of a nonfinal coordinate clause, whereas kanasi-ku is the first half of a complex predicate. In (58), tadasi-ku is the first half of a complex attribute. The infinitive is one of the most important of the OJ adjective forms. It is also found in the modern standard language with largely unchanged functions (although it has been subject to phonological change in EMJ, see 8.2). (53)

minatwokaze

samu-ku

puku

rasi

rivermouth.wind cold-ACOP.INF blow.CONCL PRES 'the wind from the rivermouth seems to blow coldly' (MYS 17.4018) (54)

topo

no

kuni

distant COP.ADN country

yamato

wo

mo

ni

imada

DAT

yet

topo-ku

mo ETOP

tuka-zu arrive-NEG.INF

sakarite

Yamato ACC ETOP distant-ACOP.INF separated.from.GER 'not yet having reached that distant land and also far separated from Yamato' (MYS 15.3688)

86

3 OJ grammar

(55)

kimi ga yuki ke naga-ku nari-nu my.lord GEN go.INF day long-ACOP.INF become-PERF.CONCL 'many days have passed since you, my lord, left' (KK 88)

(56)

wo naka ni penarite topo-ku yama-kapa mountain-river ACC middle DAT obstruct.GER distant-ACOP.INF tomo kokoro wo tika-ku omopose although heart ACC close-ACOP.INF think.RESP.IMP wagimo my. beloved 'even if we are far apart, with mountains and rivers between us, think our hearts close to one another, my love' (MYS 15.3764)

(57)

wo mireba taputwo-ku titi-papa father-mother ACC see.PROV awesome-ACOP.INF mye-kwo mireba kanasi-ku megu-si woman-child see.PROV dear-ACOP.INF lovely-ACOP.CONCL 'when one sees one's mother and father, they are awesome; when one sees one's wife and child(ren), they are dear and lovely' (MYS 18.4106)

(58)

tadasi-ku kiywo-ki true-ACOP.INF pure-ACOP.ADN 'with a true and pure heart' (SM 29)

kokoro heart

WO

ACC

motite hold.GER

The infinitive is also used as the stem upon which were built a number of forms: the gerund (taka-kute) and an extended gerund (taka-kusite, usually thought to involve the gerund of se- 'do'). The infinitive also formed the basis for the formation of the innovative conditional (taka-kupa) and it combined with the concessive conjunctional particle (topo-ku tomo, cf. (56)). Note that particles such as pa followed a finite form of verbs; this is also the case with the concessive particle tomo which followed the conclusive of verbs. Also the evidential final particle miyu which followed the conclusive of verbs is said to have followed the infinitive of adjectives, but the few examples of this rely on reading tradition and are not phonographically attested. Finally, there are a very few examples of a nominal use of the infinitive, e.g. tokizi-ku adnominalized with no: tokizi-ku no 'perennial'; and adverbialized with ni: tokizi-ku ni 'perennially'.

3.2.2.4.2 Infinitive-2 -mi is not usually termed infinitive. While it is functionally more limited than the verbal infinitive, those functions which it has are similar to those exhibited by the verbal infinitive or gerund. Functionally it

3.2 Adjectives

87

often corresponds to the infinitive-} in -ku. In later stages of the language, -mi came to be used as a concrete nominalizer, but this was very rare in OJ, a singular example being sige-mi 'thicket' (sige- 'dense, thickly growing'). In OJ, infinitive-2 was a subordinate, predicative form, that is, an adverbial, nonfinite form. Its main use was in free adverbial clauses, usually of the form NOUN (wo) ADJ-mi. It is often interpreted to mean 'as, because', but the basic meaning is simply 'being'. (59)

yama wo taka-mi (MYS 1.44) yama taka-mi (MYS 3.324) yama-daka-mi (KK 78) 'the mountain being high; as, because the mountain is high'

(60)

opokimi no kokoro wo yura-mi ami no kwo great.lord GEN heart ACC slack-ACOP.INF Omi GEN child no yapye no siba-kaki iritata-zu ari GEN manyfold COP.ADN twig-fence enter-NEG.INF exist.CONCL 'the heart of the great lord being slack, he does not enter the manyfold twig fence of the young Omi' (KK 107)

Infinitive-2 was also used as a complement, i.e. as the predicate in governed adverbial clauses. This use is frequent in Senmyo (and in kanbun-kundoku), but is found also in the poetry. Used with omop- (or a synonym) this means 'find, deem NOUN to be ADJ'. Used with se- the meaning is 'treat NOUN as ADJ; find NOUN ADJ'. This pattern was grammaticalized and has over time yielded a number of lexicalized forms which survive into the modern language, e.g. omonze-lomonzi- 'value, give weight to' (< omo-mi se-; ama'heavy'), karonze-!karonzi- 'make light of' (< karo-mi se-; EMJ karo- 'light'). (61)

ne-siku wo . . . urupasi-mi omopu sleep-SPST.NMNL ACC wonderful-ACOP.INF think. CONCL 'I think it wonderful that she slept (with me)' (KK 46)

(62)

ima no masaka mo urupasi-mi sure now COP.ADN this.moment ETOP wonderful-ACOP.INF do.EXCL 'I find also this very moment lovely!' (MYS 18.4088)

The formant -mi is often regarded as a fossilized infinitive of a verbal derivational morpheme, *-m-. Support for this may be found in singular examples such as kasikwo-mi-te (NSK 102) 'reverently, with reverence' (kasikwo- 'be awe-inspired') which has -mi with the verbal gerund formant -te, or yorokobwi

88

3 OJ grammar

yorosimi 'rejoicing and being glad' (EN 26) where the adjectival infinitive yorosi-mi seems to be parallel with the verb infinitive yorokobwi. It should be noted, however, that -mi also occurs in morpho-syntactic contexts which are unusual for a verbal infinitive. Interestingly, the morpho-syntax of -mi is to a large extent shared by desiderative pori-, the infinitive of a defective verb por- 'wish, want, love', and by -ni, the infinitive of the negative auxiliary -(a)n-. por- is found mainly in a few fixed expressions. It takes a nominal complement, either a noun (usually me 'eye' in the set phrase N game wo por- 'I want to seeN') or a verb in the nominal form (usually V-maku (wo) por-, i.e. the nominal of the conjectural-(a)m- 'I want to V'). -ni and pori may both be used as free adverbials and to complement omopu. (63)

mi-maku pori nisi no mimaya no see-CONJ.NMNL want.INF west COP.INF stable GEN ni tat-era-masi DAT stand-STAT-SUBJ.CONCL 'wanting to see my beloved I would be standing outside the western stables (Umaryo)' (MYS 15.3776)

two door

(64)

wa ga kokoda sinwopaku sira-ni pototogisu I GEN this. much long.NMNL know-NEG.INF cuckoo idu-pye no yam a WO ka naki COP.ADN mountain ACC cry.INF which-direction Q kwoyu ramu PCONJ.ADN cross.over.CONCL 'over which mountain will the cuckoo fly crying, unaware that I long so much' (MYS 19.4195)

(65)

mi-maku pori omopu see-CONJ.NMNL want.INF think..CONCL 'I am thinking that I want to meet/see you' (MYS 17.3957)

(66)

mo yuki-sugwi-kate-ni omop-yereba inabinwo ETOP go-pass-be.possible-NEG.INF think-STAT.PROV Inabino 'as I was thinking that it is not possible to leave Inabino' (MYS 3.253)

-mi, pori and -(a)ni may be further adverbialized with to, i.e. they form a slightly irregular gerund in -to instead of -te. (67)

tukapi no kyereba uresi-mito messenger GEN come.STAT.PROV happy-ACOP.GER '(being) happy because a messenger had come' (MYS 17.3957)

3.2 Adjectives

89

(68)

kurapasi-yama wo sagasi-mito Kurahashi-mountain ACC steep.ACOP.GER 'Mt. Kurahashi being steep' (KK 69)

(69)

naku kowe wo kika-maku porito sing.ADN voice ACC hear-CONJ.NMNL want.GER 'wanting to hear the crying voice (of the cuckoo)' (MYS 19.4209)

(70)

aka-nito be.satisfied-NEG.GER 'without being satisfied' (MYS 17.3991)

-mi and pori are generally not used in other inflected forms. In addition to kasikwo-mite mentioned above, the exceptions, which are very few, include singular instances of adnominal poru (a ga poru tama (NSK 92) 'the pearl that I want/love') and pori with -si the simple past adnominal (wa ga pori-si arne pa puri-ki-nu (MYS 18.4124) 'the rain I wished/longed for has started to fall'). Like -mi, both -ni and pori could be predicated by se- 'do'. For pori this construction was in EMJ phonologically reduced to give the desiderative verb posse- <pori-se-. 9 Note also the coradical shiku adjective OJ posi 'be desirous of' which may also be from pori-si.

(71)

ono ga inoti wo naga- ku pon own GEN life ACC long-ACOP.INF want.INF 'I wish for my life to be long' (MYS 12.2868)

(72)

kimi pa mire do aka-ni se-mu my.lord see.CONC be.satisfied-NEG.INF do-CONJ.CONCL 'although you look at it, my lord, you will not be content' (MYS 17.3902)

sure do.EXCL

Finally, -mi and pori are used with the particle kamo (otherwise used with the adnominal of verbs), expressing doubt or exclamation (-(a)n-, however, uses the adnominal with kamo, cf. (74)). (73)

9

pototogisu naku oto paruke-si cuckoo cry.ADN sound distant-ACOP.CONCL satwo-dopo-mi kamo village-distant-ACOP.INF Q 'the sound of the cuckoo crying is distant, maybe because the village is far away' (MYS 17.3988)

The only example with pori written phonographically before se- is in MYS 14.3383.

90

3 OJ grammar

(74)

kokoro so ita-lei . . . mi-maku pori kamo heart FOC painful-ACOP.ADN see-CONJ.NMNL want.INF Q 'my heart aches, maybe because I want to see you' (MYS 20.4307)

3.2.3

Analytic forms

In addition to its basic inflected forms, the adjectival copula formed extended analytic forms by the infinitive -ku followed by ar- (or another existential verb): -ku ar-. This allowed the formation of forms which were either not or only rarely formed directly in the primary paradigm, including combination with auxiliaries which did not combine directly with the adjectival copula, e.g. (75) with the modal past -(i)kyer-. The analytic formations provided full morphophonological versatility for the adjectival copula. The analytic forms sometimes fused phonologically, -ku ar- => -kar-, which in EMJ gave rise to a secondary conjugation of the adjectival copula; see further 3.4.2.1.1 and 3.4.2.1.2. (75)

kimi ga yosopi si taputwo-ku admirable-ACOP.INF my.lord GEN attire EMPH ari-kyeri exist-MPST. CONCL 'your attire, it is admirable, my lord!' (KK 7, NSK 6)

3.2.4

Ku versus shiku a4Jectives

Shiku and ku adjectives cannot be formally distinguished on the basis of the conclusive form, which does not make clear whether the final /si/ is part of the stem or is the conclusive form of the adjectival copula, cf. conclusive yosi 'good' and asi 'bad'. However, adnominalyoki and asiki show that the stems are yo- and asi- respectively. Unger and Tomita (1983) point out that on the criterion of phonographic attestation of a form other than the conclusive/stem, it is actually only possible to determine positively the conjugational class of approximately two thirds of all OJ adjectives. There are a number of points concerning the shiku adjectives, especially their origin and original properties, which remain unclear, but two tendencies in the semantic and morphological properties of the members of each class are notable: First, semantically, most shiku adjectives are 'psych' adjectives, referring to subjective emotional states, whereas the ku adjectives typically express more objective qualities. This was originally observed by Yamamoto (1955). On his calculations there are approximately 20 per cent exceptions to this tendency in OJ (12% for ku, 26% for shiku) and more exceptions in later

3.2 Adjectives

91

stages. (76) gives for each class a larger group confonning to the tendency and a smaller group of exceptions. Second, morphologically, the shiku adjectives may generally be thought to involve a fonnant -si (or possibly -Vsi). It should be noted, however, that in not a few cases it is not possible to isolate a stem for the derivation which may be identified with an otherwise known morpheme. It is thus possible that some shiku adjectives have simple stems which happen to end in si; for example, asi- 'bad' or wosi- 'dear' could well be simple stems. On the other hand, there are no ku adjectives with stem final /-i/, and it remains likely that all adjective stems ending in /si/ are in fact derived. It has been noted that a large group of shiku adjectives are transparent deverbal derivatives (Yamazaki (1992)); this is shown in (76c) which also gives the verbs that Yamazaki posits as derivational stems for such adjectives. This relation is quite different from that holding between some ku adjectives and coradical verbs, e.g. aka- 'bright, red', ake- 'become bright', akas- 'make bright'; ara- 'rough', are- 'rage; get ruined', aras- 'damage, ruin', where the verbs and the adjective originate in a common (nominal) root. It is on the other hand clear that not all shiku adjectives are deverbal. There is no consensus about the origin or morphological status or function of the fonnant -si. Some scholars (including Yamamoto 1955, Yamazaki 1992) believe that it reflects a derivational fonnant -si (or -Vsz) etymologically different from the conclusive fonnant -si; varying degrees of functional specificity are posited (adjective formant, psych adjective fonnant, deverbal psych adjective fonnant). The other main point of view, represented by e.g. Kawabata (1976) is that -si is etymologically identical with the conclusive fonnant of the ku-adjectives and that this -si in the course of the fonnation of the adjective inflection was resegmented as a part of some adjectives (shiku), but not of others (ku). On that view the semantic specialization between ku and shiku adjectives and the function of -si as a derivational fonnant are secondary and would be a result of a further reanalysis of the shiku adjectives as consisting of a stem and a derivational suffix. Regardless of its etymology, it seems that the synchronic function of OJ -si was simply to derive adjectives from other parts of speech, without any semantic specification. It may be viewed as having arisen in the course of the establishment of adjectives as an independent part of speech. Identifying -(V)si as the carrier of the psychological meaning of the shiku adjectives is problematic. First, there is the not insignificant number of exceptions (26%). Second, the psychological meaning may in many cases be seen to reside in the lexical semantics of the verbal stem. Third, there may well have been a tendency for adjectives derived by means of -si to specialize semantically without that being a feature of the meaning of -si: adnominalizing

92

3 OJ

grammar

or adverbalizing verbs or nouns was morphologically unproblematical and a derived adjective would therefore only be derived if some special aspect of or perspective on the semantics of the stem was required. (76) a.

b.

c.

d.

Ku adjectives aka- 'bright, red'; ara- 'rough'; asa- 'shallow'; kata- 'hard, firm'; kurwo- 'black'; mane- 'frequent'; na- 'non-existent, no'; naga'long'; nuru- 'tepid'; opo- 'many, plentiful';paya- 'fast';puru'old';putwo- 'thick, sturdy'; siru- 'obvious, as may be expected'; sirwo- 'white'; tika- 'near'; topo- 'distant'; usu- 'thin, weak'; waka- 'young';yasu- 'peaceful, easy'; yo- 'good' ita- 'painful'; kasikwo- 'fearsome, awesome'; niku- 'disagreeable'; tayu- 'exhausted'; u- 'sad' Shiku adjectives asi- 'bad, evil'; atarasi- 'precious, regrettable, dear' (atar- 'touch'); kanasi- 'dear, sad' (kane- 'be unable to'); kokidasi- 'grave, serious'; kuyasi- 'regrettable, vexing' (ku(y)i- 'regret'); kwopwisi-1 kwoposi- 'dear, beloved' (kwopwi- 'love'); kyesi- 'strange, unusual'; natukasi- 'dear, yearned for' (natuk- 'become familiar with, be fond of'); opoposi- 'dim, gloomy';pasi- 'beloved';posi- 'desirable' (cf. por- 'want', see 3.2.2.4.2); sabusi-/sabisi- 'sad, lonely' (sabwi- 'get desolate'); tagitagisi- 'uneven'; tomosi- 'sparse, enviable, poor'; uramyesi'regrettable' (*urami- 10 'regret, resent'); uresi- 'joyous'; urupasi'beautiful' (urup- 'get wet'); wabwisi- 'lonely' (wabwi- 'be embarrassed'); wemapasi- 'likeable, smile-provoking' (wemap- 'keep smiling' <= wem- 'smile'); wosi- 'dear'; yorosi- 'suitable' (yor- 'approach'); yuyusi- 'awesome' ikasi- 'plentiful, vigorous'; iyasi- 'lowly'; munasi- 'empty'; pisasi- 'long-lasting'; sagasi- 'steep'

3.2.4.1 Reduplication A final point about shiku adjectives: traditionally, reduplicated adjectives are said to belong to shiku, cf. tagitagisi- (*tagi- '?')'uneven' and perhapsyuyusi(*yu- '?')'awesome'; this is certainly the case from EMJ onwards. While this was in all likelihood also the case in OJ, it should be noted that OJ reduplicated 10

urami- is not phonographically attested in OJ. This verb was UB in EMJ, but may have been ura-mi- (UM) 'heart-see' in OJ.

3.3 Copula

93

adjectives with identifiable stems are only attested in the conclusive, making it impossible to determine whether they really are shiku adjectives, e.g. naganagasi 'very long' (naga- 'long'; EMJ naganagasi-), topotoposi 'very far, far far away' (topo- 'distant'; EMJ topodoposi-), wowosi 'gallant' (wo 'male'; EMJ wowosi-). 3.2.4.2 Jiku aqjectives The jiku adjectives are usually regarded as a subclass of shiku adjectives. It is a very small closed class. (77) is a full list of jiku adjectives. With the exception of masizi-, the OJ jiku adjectives are fairly transparent denominal derivatives with a formant -zi and have the meaning 'like NOUN, typical of NOUN'. It may be noted thatkono ipyeziku mo 'also like this house' (SM25, the only attestation of ipyezi-) has the noun phrase kono ipye as the derivational stem for -zi.

(77)

ipyezi- 'like (this) house' (ipye 'house'); masizi- 'negative potential'; omozi- 'close (as if from same mother)' (omo 'mother'); onazi-- oyazi- 'same, similar' (cf. ono 'self'); tokizi- 'timeless, eternal, constant; untimely' (cf. toki 'time'); warezi- 'like me' (ware '1'). From EMJ: imizi- and susamazi- (- susamasi-)

Note in this connection the OJ construction NOUN-zi mono 'like (a) NOUN, typical of (a) NOUN, as befits (a) NOUN' (e.g. uma-zi mono 'like a horse'), built on animate or concrete nouns (in addition to a single example with the stem of a ku-adjective: kasikwo-zi mono (SM 14) 'reverently' (kasikwo- 'fearsome, awesome')). The construction is attested with the following nouns: inu 'dog', i 'cormorant', uma 'horse', kakwo 'fawn', kamwo 'duck', sisi 'wild animal', toko 'bed', tori 'bird', wotokwo 'man',yuki 'snow'. Thejiku adjectives and this construction involve a derivational suffix -zi < *-nVsi, see 3.5.2. Further, masizi and tokizi 'untimely, timeless' are usually said to involve a different derivational -zi which is negative in meaning and possibly related to the negative auxiliary -(a)zu, or the negative conjectural flective -(a)zi. 3.3

Copula

OJ had two simple copulas with highly defective inflection. The two copulas are usually not thought of as verbs, but functionally they exhibit the forms in Table 3.9. Morphologically, infinitive ni, gerund nite and adnominal tu look like regular verb forms, whereas no and to appear to be roots used in adnominal and adverbial function.

94

3 OJ grammar Table 3.9 OJ copula forms

Adnominal Infinitive Gerund

OJ

EMJ

(tu) no ni-no to nite

no ni to nite >de

The basic function of the copula in Japanese is to predicate, adnominalize or adverbialize nouns. As we saw above (3.2) that also, to a limited extent, included adjective stems in OJ. As the simple inflected copulas were highly defective, analytic constructions with a copula infinitive or gerund followed by an existential verb were used in many functions (3.3.2). All the copula forms in Table 3.9 were also used as case and other particles, see 3.7.8.2.1. With the exception of the loss of tu, and the change of nite > de, they are all still in use today in NJ both as basic copula forms and as case particles, constituting a remarkably stable pervasive feature of Japanese through its attested history, with little change in form or function. 3.3.1

Use of the simple iriflected copula forms

Of the two adnominal forms, tu was already in OJ restricted and used only to adnominalize nouns or adjective stems directly, and it disappeared in HMJ outside lexicalized expressions~ it is usually described as a genitive, cf. 3. 7.1. As in NJ, no was highly frequent in OJ and was in addition to the marginal use with adjectives (see (38c) above) used in adnominal clauses as in (78). As distinct from the adnominal forms of verbs and auxiliaries, no could not be used as a nominalized form and it was not used as the predicate in kakarimusubi constructions (see 8.9)~ for those purposes, analytic forms were used (see 3.3.2). (78)

purupye no satwo no asuka old.house GEN village COP.ADN Asuka 'Asuka which is the village of my old home' (MYS 3.268)

The non-finite copula forms were used in subordinate function 'being, as, like', (79)~ infinitive no is rare and only used in this sense. Infinitive ni and to could also be used simply to adverbialize, see (80), and in small clause constructions,

3.3 Copula

95

see (81), similar to the accusative-with-infinitive construction in some other languages. (79) a.

purite namida kwosame ni tear small.rain COP.INF fall.GER 'tears fall like fine rain' (MYS 2.230)

b. pana no . . . yuki to puri-kye-mu flower GEN snow COP.INF fall-SPST-CONJ.ADN 'flowers would fall like snow' (MYS 17.3906) c.

asa-pi no wemi-sakaye-kite morning-sun COP.INF smile-prosper-come.GER 'come smiling radiantly as the morning sun' (KK 3)

(80)

wotomyera no kazasi no tame ni ladies GEN hairpin GEN sake COP .INF miyabwiwo no kadura no tame to elegant.fellow GEN headdress GEN sake COP.INF 'for the sake ofthe hairpins of the ladies, for the sake of the headdresses of the elegant fellows' (MYS 8.1429)

(81) a.

terasu pi wo yam1 ni minasite shine sun ACC darkness COP.INF see/regard.GER 'seeing the shining sun as darkness' (MYS 4.690)

b. kagami no yama wo miya to sadamuru Kagami GEN mountain ACC residence COP.INF decide.ADN 'decide on Mt. Kagami as his residence' (MYS 3.417)

3. 3.2

Analytic forms

Supplementing the few simple inflected copula forms were analytic forms consisting of an infinitive (ni, to) followed by ar-, e.g. (82a-b), or another existential verb, e.g. (82c), sometimes separated by a particle, e.g. (82d). Analytic forms were the only way of forming most inflected forms for the copula and for combination with tense and mood auxiliaries and with extensions, and the analytic formations were thus an indispensable part of the morphology and syntax of the copula. Ni ar- sometimes phonologically fused to give nar-, see (82e), which in EMJ became a fully inflected copula (see further 3.4.2.1.1-2):

96

3 OJ grammar

(82) a.

ima koso pa wa-dori now FOC TOP me-bird 'Now, I am my own' (KK 3)

b.

yononaka this.world

pa TOP

ni

ara-me

COP.INF

exist-CONJ.EXCL

munasi-ki empty-ACOP.ADN

mono thing

to

COP.INF

ara-mu exist-CONJ.CONCL 'this world is empty' (MYS 3.442) c.

3.4

na koso pa wo you FOC TOP man 'as you are a man' (KK 5)

ni

imaseba

COP.INF

exist.RESP .PROV

d.

m1s1pye no nana no sakasi-ki pitodomo past GEN seven GEN wise-ACOP .ADN people mo pon-se-s1 mono pa FOC want-do-SPST.ADN thing TOP sake ni si aru rasi sake COP EMPH exist.ADN PRES 'It seems that what the seven wise men of old, too, wanted is sake' (MYS 3.340)

e.

ume no pana ima sakari nari plum GEN blossom now in.bloom COP.CONCL 'The plum blossom is in bloom now' (MYS 5.850)

Conjugation classes and morphophonology

OJ has eight verbal conjugation classes. Although not usually included as verbal conjugation classes, the simple past auxiliary (3.1.4.2), the negative auxiliary (3.1.4.2), and the adjectival copula (3.2) have irregular conjugations with forms for most of the categories which verbs inflect for. Further, as mentioned above, the copulas (3.3) may also be considered verbs with highly defective and irregular conjugations, and finally, the infrequent auxiliary verb -kose- (3.4.3.3) has a defective and slightly irregular conjugation. From a synchronic morphophonological point of view, the eight OJ verbal conjugation classes fall in two major groups, consonant base and vowel base verbs, each with regular and irregular subclasses. The names for the verb classes are explained in 3.4.6.2. There are two major lexical verbal classes: quadrigrade (abbreviated QD), which holds approximately 75 per cent of all OJ verbs, and lower bigrade (LB), which accounts for approximately 20 per cent of the verbs. The remaining classes have small sets of members. There are around thirty upper bigrade (UB) verbs and fewer than a dozen upper mono-

97

3.4 Conjugation classes and morphophonology

grade (UM) verbs. n-irregular (n-irr) holds two lexical verbs and an auxiliary. r-irregular (r-irr), k-irregular (k-irr), and s-irregular (s-irr) are each defined by a single idiosyncratic and highly frequent verb with some amount of grammatical functions: ar- 'be, exist', ko- 'come', and se- 'do', respectively. (83)

Consonant base Regular Quadrigrade Irregular r- irregular n- irregular Vowel base Regular Lower bigrade Upper bigrade Irregular

3.4.1

Upper monograde k- irregular s- irregular

75% of OJ verbs

kak-, or-, ... arsin-, in-, -(i)n-

20% of OJ verbs c. 30 verbs

ake-, ware-, ... okwi-, ori-, ...

c. 10 verbs

mi-, ni-, . .. kose-

Regular vero classes

3.4.1.1 Quadrigrade verbs QD is the class of regular consonant base verbs and the largest verb class. The base ends in a consonant and most have the shape CVC-. Those which are longer, CVCVC-, usually incorporate more or less transparent derivational matter. The following base final consonants are found: -p, -t, -k, -b, -g, -m, -s, -r. Thus, there are no regular consonant base verbs which end in -d, -n, -z, -y, -w. 3.4.1.2 Bigrade verbs LB is the large regular class of vowel base verbs, with bases ending in -e (that is, -e 2, as opposed to -e 1 after consonants where the distinction existed). Most LB verb bases are disyllabic, CVCe-, but there are a few monosyllabic, Ce- (e.g. e- 'get', pe- 'pass', ne- 'sleep'), and some trisyllabic CVCVCe- bases as well. UB verbs end in -wi (neutralized as -i after alveolars). They number only about thirty. Most are disyllabic, CVCwi-, but a few monosyllabic, Cwi-, and trisyllabic, CVCVCwi-, bases exist. The few CVi- bases (koi- 'lie down', kui- 'regret', oi- 'age, get old') are underlyingly //CVyi-// as shown both by derived forms such as kuyasi- 'regrettable' and by their EMJ conclusive forms (they are only attested in their base in OJ). (84) is a reasonably full list ofUB verbs in OJ and MJ. Note that some verbs are attested only in forms that do not allow an unambiguous identification of their conjugational class in OJ. For example, for osori- 'fear' only the base (/infinitive) is attested in OJ, osori, and conversely porobwi- 'go to ruin' is only attested in the conclusive, porobu;

98

3 OJ grammar

both of these verbs could therefore possibly belong to QD in OJ Where the attested OJ forms of a verb do not allow us positively to identify the conjugational class, we here project the EMJ class back to OJ, but note relevant facts of attestation. (84)

koi- 'lie down'[= //koyi//; only base attested in OJ] komwi- 'enclose, put in' [only attested once in the compound tumagomwi 'wife-enclosing'; usually LB kome-] kozi- 'uproot' [only base attested in OJ] kui- 'regret' [= /!kuyi//] kwopwi- 'love' nagwi- 'become calm' negwi- 'appease, solace, pray, be sympathetic with' [only base attested in OJ] odi- 'fear' [only conclusive attested in OJ] oi- 'age, get old'[= //oyi//; only base attested in OJ] okwi- 'arise' opwi- 'grow/get bigger' osori- 'fear, dread, be apprehensive' [only base attested in OJ] oti- 'fall' pwi- 'winnow' [only one OJ example, in the compoundpwi-gapa (place name)] porobwi- 'go to ruin' [only conclusive attested in OJ] puri- 'get old' [only base attested in OJ] pwi- 'dry (intr.)' pwi- 'sneeze' sabwi- 'get desolate, old, rusty, faded' sakwi- '(?)be tom' [rare; only base attested in OJ; usually LB sake- 'be torn'] sinobwi- 'endure, suffer, hide' sipwi- 'insist, force' sugwi- 'pass by, exceed, elapse, pass away' susabwi- 'become strong' [from EMJ also susab- and susam-, both QD] tamwi- 'go round, turn' todomwi- 'stop (tr.)' [also LB todome- which is more common] tukwi- 'get exhausted' wabwi- 'be embarrassed, disappointed, apologize for' wi- 'sit down, settle down' [wi-, which is the ancestor ofNJ i- 'exist', is often treated as an existential verb also in OJ, but it was not used as a simple existential verb until late LMJ, see 12.4]. woti- 'be rejuvenated' [only base attested in OJ]

3.4 Conjugation classes and morphophonology

99

yodi- 'grab and pull' [only base attested in OJ] yokwi- 'avoid, go around' . . . . -bwi- 'be like ... '; derives verbs from nouns or adjectives, e.g.: arabwi- 'be wild' (ara- 'wild'), manabwi- 'imitate; learn' manebwi- 'imitate' (cf. mane 'imitation'). ... -sabwi- 'act as/display characteristics appropriate for ... '; derives verbs from nouns, e.g.: kamusabwi- 'act godly, be old' (kamu-!kamwi 'spirit, god'), yamasabwi- 'be mountain-like' (yama 'mountain').

Additional UB verbs (attested in EMJ unless noted): (85)

abi-lami- 'bathe oneself in' iki- 'live' [from LMJ; QD in OJ and EMJ] isati- 'weep' [attested as UM in OJ] kabi- 'tum mouldy' kobi- 'flatter' koi- 'freeze' [= //koyi/1] kori- 'feel regret/remorse for, learn by experience' kuti- 'rot' miti- 'become full' [from LMJ; QD in OJ and EMJ] mukui- 'requite'[= //mukuyil/] nebi- 'get, seem old' nedi- 'twist, screw' nobi- 'extend, lengthen, spread, grow' obi- 'tie around the waist' [QD in OJ] ori- 'descend, alight' padi- 'feel shame' potobi- 'swell with water' simi- 'freeze' tibi- 'get worn out' [not attested until LMJ] todi- 'shut' urami- 'resent, regret'

3.4.1.3 Formation ofbasic inflectedforms There are five basic inflected verb forms (see Table 3.10). Their formation on the regular verbs can synchronically be described in terms of combination of the verb base with a flective, with reference to simple phonological rules, in particular the regular rules of vowel deletion which apply when two vowels are brought together in composition or morphological derivation, see 2.6.1. We will briefly discuss alternative analyses which have been proposed.

100

3 OJ grammar

Table 3.10 Formation of OJ inflected verb forms

QD Base

kak-

war- 'break (tr.)'

Infinitive Imperative

kaki <= kak + i kakye <= kak + ~

wari <= war + i ware <= war + ye

Conclusive Adnominal Exclamatory

kaku <= kak + u kaku <= kak + ru kake <= kak + re

waru <= war + u waru <= war + ru ware <= war + re

Base

like-

ware- 'break (intr.)

e- 'get'

Infinitive Imperative

ake+ 0 ake (+yo)

ware+ 0 ware(+ yo)

e+0 e +yo

Conclusive Adnominal Exclamatory

aku <= ake + u akuru <= aku + ru akure <= aku + re

waru <= ware + u waruru <= waru + ru warure <= waru + re

u <= e + u uru <= u + ru ure <= u + re

Base

okwi-

ori- 'descend'

pwi- 'dry'

Infinitive Imperative

okwi + 0 okwi +yo

ori+ 0 ori +yo

pwi+0 pwi +yo

Conclusive Adnominal Exclamatory

oku <= *okwu <= okwi + u okuru <= oku + ru okure <= oku + re

oru<= ori + u oruru <= oru + ru orure <= oru + re

pu <= "jJwu <= pwi + u puru <= pu + ru pure <= pu + re

LB

UB

The conclusive, adnominal and exclamatory endings are common to the two conjugational classes. The following regular phonological rules apply in the formation of these three forms. The important point about the vowel deletion rules is that when two vowels come together, the second is only deleted in the context -(C)V-VCV-, i.e. when a monosyllabic morpheme is followed by a disyllabic vowel initial morpheme; in all other cases, it is the first vowel which is deleted (cf. 2.6.1): (86)

Vowel deletion ake + u => aku okwi + u => *okwu => oku e + u => u Consonant deletion kak + ru => kaku kak + re => kake

3.4 Conjugation classes and morphophonology

101

The vowel base verbs attach the adnominal and exclamatory endings to the conclusive, rather than to the base. The conclusive is thus used as a stem in addition to its use as an inflected form. An alternative analysis would posit the adnominal and exclamatory formants as -unt and -ure, respectively, for the vowel base verbs. This would mean, first, that the flectives are different for the consonant and vowel stem verbs. Second, and more importantly, specific morphophonemic rules would be required to generate the correct forms, for as shown in (87b), the application of the regular phonological rules would generate incorrect forms for the monosyllabic vowel bases. The proposed analysis gives the correct forms in all cases, (87a): (87) UB CVCwi LB CVCe UBCwi LBCe

a. Conclusive + -nt (I-re) oku + nt => okunt 'rise'

b. Base + -unt (l-ure) okwi + unt => (okwunt =>) okunt ake + unt => akunt aku + nt => akunt 'open' pu + nt =>punt 'dry (intr.)' pwi + unt => *pwiru pu + nt =>punt 'pass, elapse' pe + unt => *peru

The consonant base verbs form their infinitive and the imperative by means of flectives, -i and -ye, respectively, whereas the vowel base verbs simply use the base as infinitive and, originally, imperative (see further 3.4.5). The infinitive formant is sometimes said to be -i also for the vowel base verbs. However, that, again, would mean that otherwise unmotivated morphophonemic rules would have to be posited, for the regular phonological rules give incorrect forms for the LB verbs, (88b): (88)

a. Base = infinitive UB CVCwi okwi UB Cwi pwi LB Ce pe LB CVCe ake

3.4.2

Irregular consonant base verbs

b. Base+ -i okwi + i => okwi pwi + i => pwi pe + i =>*pi ake + i => *aki

3.4.2.1 r-irregular r-irr is the class of irregular -r base verbs. It is defined by the highly frequent existential verb ar- and in addition also holds a number of frequent and important grammatical morphemes which are all thought to incorporate ar- etymologically. r-irr also includes the verb war- 'be sitting'. War- is the ancestor of the NJ existential verb or- and usually war- is said to have been an existential verb also in OJ However, it has been shown by Kinsui (2006) that war- was not used as a simple existential verb until LMJ. In fact,

3 OJ grammar

102

wor- was used as a lexicalized stative form of (an ancestor of) wi- 'sit/settle down', in OJ and EMJ, meaning 'be sitting' and for example as the antonym of tat-er- 'be standing' (the morphological stative of tat- 'stand up'). Note also that while wor- had grammatical uses from OJ- as an auxiliary verb to form progressives (cf. 3.1.4.7.4) - it was not used in the grammatical functions exhibited by truly existential verbs until late LMJ. See further 12.4 about wor-. r-irr also comprises two existential verbs pab(y)er- and imasugar- which are both thought to involve developments of something + ar-, although it is not clear what that something might be; neither is phonographically attested in OJ.

(89)

r-irr verbs ar- 'be; exist' wor- 'be sitting' From EMJ: pab(y)er-- panb(y)er- 'exist.humble; exist. polite' (no phonographic OJ attestation, but is found in the reading tradition of{~ which otherwise represents words meaning 'serve'). imasugar-- imasukar- 'exist.RESP' (variant shapes: imasogar-, imasokar-, imasigar-, imasikar-, mimasugar-, mimasukar-).

Other members of the r-irr class are thought to incorporate ar-, in most cases conspicuously so. The forms in (90) are fully lexicalized grammatical morphemes, etymologically incorporating ar-. The forms in (91) are phonologically fused, but not yet lexicalized forms (some of which are not attested until EMJ in their fused form); the forms in (91a) and (91b) became important grammatical morphemes in EMJ, see 3.4.2.1.2 immediately below. (90)

Lexicalized grammatical morphemes Auxiliaries -kyer- modal past< *ki-ar- 'come-exist' or 'sPST-exist' -yer- stative < *-i-ar- 'INF-exist' Evidential verb extensions nar- 'sounds' < *na-ar- 'sound-exist' From EMJ: mer- [not robustly attested until EMJ] 'looks'< *me-ar'eye-exist' or *mi-ar- 'see-exist'

(91) a.

Phonologically fused forms Periphrastic stative (gerund + ar-) -tar- <= -te ar-

3.4 Conjugation classes and morphophonology

b.

Infinitive + arnar- <= ni ar- 'COP.INF +exist' tar- [not until EMJ] <=to ar- 'cOP.INF +exist' kar- <= ku ar- 'ACOP.INF +exist' be-kar- <= be-ku ar- 'NEC-ACOP.INF + exist' zar- <= zu ar- 'NEG.INF + exist'

c.

Adverb + arkakar- 'be this way'<= kaku ar- 'this.way +exist' sikar- 'be that way'<= sika ar- 'that.way +exist' sar- 'be that way' [not until EMJ] <=saar- 'that.way +exist'

103

r-irr verbs can be said to be syntactically rather than morphologically irregular: they use the infinitive as conclusive form. In other respects they are like the QD verbs. Note that the verb extensions be-, rasi-, and ram-, which regularly attach to the conclusive, attach to the adnominal of r-irr verbs: aru be-, aru rasi-, aru ram-. Another way of describing this would be to say that r-irr verbs have two conclusive forms: regularly formed aru (<= ar+u) used only with verb extensions, and ari (=infinitive) used elsewhere.

(92) Infinitive Imperative

arari <= ar+ i are<= ar+ ye

Conclusive Adnominal Exclamatory

ari <= ar+ i aru <= ar+ ru are<= ar+ re

3.4.2.1.1 Grammatical uses of ar- and other existential verbs In addition to its lexical meaning, the existential verb ar- has been used extensively in grammatical functions throughout the history of Japanese. Thus, some important grammatical morphemes which were lexicalized already in OJ etymologically incorporate ar-, see (90) above. However, productively ar- - and other existential verbs, in particular the exalted synonyms of ar- (e.g. in OJ respectful mas-, imas-) -has been used to form extended, analytic or periphrastic forms through all attested stages of Japanese (reflected in (91) above), and these uses of existential verbs are a pervasive feature of Japanese. There are two overall uses: First, reflecting its semantics as a stative verb, ar- formed stative verb forms, both in the morphological stative (-yer- < *-i ar-) and in the periphrastic stative construction (-tear-), see 3.1.4.7.3. Second, used with infinitives of the adjective auxiliary (-ku), negative auxiliary (-zu) or the defective copula verbs (ni, to), ar- functioned to provide analytic, morphologically versatile forms; such extension was in itself semantically vacuous (and seems

104

3 OJ grammar

to have been applied playfully in the poetic texts on occasion), but for adjectives and the negative it provided freer combination with auxiliaries (see 3.1.4.8 and 3.2.2.4.1), and for the copula it was the only way offorming most inflected forms for use in nominal predications (see 3.3.2 above). Such constructions also made it possible to focus an adjectival, negative, or nominal predicate with a focus particle after the infinitive (cf. 8.9). (93) exemplifies a focused negative (mi-zu so aru) and also has the combination of an adjective (here the extension be-) with a tense auxiliary (be-ku ari-kyeru): (93)

imo wo-ba mi-zu so aru be-ku beloved ACC-TOP see-NEG.INF FOC exist NEC-ACOP.INF ari-kyeru exist-MPST.ADN 'I should never have met my beloved' (MYS 15.3739)

3.4.2.1.2 Fused forms; secondary conjugations Periphrastic and analytic forms with ar- sometimes fused as summarized in (94): (94) a.

Periphrastic stative

b. Copula Negative Adjective

-tear-=>

-tar-

ni ar- => -zu ar- => -ku ar- =>

nar-zar-kar-

In EMJ these phonologically fused forms were lexicalized to form a new stative auxiliary, -(i)tar- (see 8.4.2), and secondary conjugations of the copula, adjective, and negative auxiliary (see 8.2.1), but in OJ they were simply fused forms produced according to the rules of vowel deletion (cf. 2.6.1). The fused forms are concentrated in the younger sections of the Man 'yoshii (and in the Bussokuseki-ka), but are not found in the early OJ materials, 11 and the unabbreviated forms were much more frequent than the fused forms. The distribution of fused and full forms in the OJ texts is sometimes taken to indicate that the phonological fusion itself was of fairly recent origin. However, this seems rather to be a matter of writing: contrast the following examples, with taputwo'awe-inspiring' from the Kojiki and Nihon shoki, (95) below, and from the Bussokuseki-ka, (96) below. Although the Kojiki and Nihon shoki form is unabbreviated, the line is metrically irregular in having eight rather than seven feet, indicating that ta.pu.two.ku.a.ri.kye.ri actually is a morphophonemic writing of taputwokarikyeri, the form phonemically written in the Bussokuseki 11

Thus, -zar- and -lear- are not found in the Kojiki, Nihon shoki, Norito, or Senmyo, and -tar-, which is attested less than fifty times in the entire OJ corpus, not in the Kojiki or Nihon shoki, only once in the Norito, and three times in the Senmyo.

3.4 Conjugation classes and morphophonology

105

-ka. Most, although far from all, examples of unabbreviated -ku ar- occur in lines with an extra written foot (so-calledji-amari 'surplus letters'), which we expect to represent a fused pronunciation. This suggests that phonological fusion was more usual than the simple ratio between written fused and full forms would otherwise lead us to believe. More importantly, unabbreviated writing of a fused pronunciation, as in (95), testifies clearly to awareness of the morphemic constituency and to the lack of lexicalization. (95)

(96)

ga yosopi si taputwo-ku ari-kyeri kimi my.lord GEN attire EMPH admirable-ACOP.INF exist-MPST.CONCL 'your attire, it is admirable, my lord!' (KK 7, NSK 6) ima

no

kusurisi

taputwo-kari-kyeri

now COP. ADN m aster.of.medicine adm irable-ACOP-MPST. CONCL 'the present master of medicine is worthy of praise!' (Bussoku 15)

3.4.2.2 n-irregular n-irr is the class of irregular n-base verbs; it has only three members, the perfective auxiliary -(i)n- and two lexical verbs sin- 'die' and in- 'depart, pass'. It is likely that this conjugation class is constituted and maintained as an independent class by the very frequent auxiliary, rather than by the two lexical verbs. Note that there are no regular n-base verbs, that is to say, no n-bases among the QD verbs and this, too, may have contributed to maintaining them orphological independence of this class. Traditional etymology posits a relation between the two lexical verbs(< *(s)in(V)-) and also derives the perfective auxiliary from in-. It is more likely, however, that the perfective -n- derives from a copula *nV, see 3.5.2. n-irr verbs have the infinitive and imperative of the consonant base verbs, but the adnominal and exclamatory are formed using the conclusive as stem, like the regular vowel base verbs. For that reason n-irr is often referred to as a 'hybrid' conjugation. Perfective -n- has no imperative. No imperative or nominal form is attested for in-; for sin-, no nominal form is found and the adnominal and exclamatory are not attested phonographically. Neither sin- nor in- forms a perfective or stative in OJ.

sin-

-n- perfective auxiliary

Infinitive Imperative

sini <= sin + i sine <= sin + ye

-ni<=-n+i

Conclusive Adnominal Exclamatory

sinu <= sin + u (sinure <= sinu + re) (sinuru <= sinu + ru)

-nu<=-n+u -nure <= -nu + re -nuru <= -nu + ru

(97)

106

3 OJ grammar

3.4.3

Irregular vowel base verbs

There are three classes of irregular vowel base verbs, all of which are monosyllabic, with bases of the shape CV-. 3.4.3.1 Upper monograde UM is the class of short (one-mora) verb bases of the shape Ci, which is thought originally to have been Ci 1. However, there are only two UM verbs which are identifiable as Ci 1 as opposed to neutral Ci (mi- 'see, look' and ki'wear'), and there is a single verb with Ci2, mwi- 'turn', which is thought to have been UB originally, partly because it has Ci2, partly on the basis of the existence of the UB verb tamwi- 'go round, turn'. Also wi- 'sit down, settle down' was UB originally, but had all but completed the shift to UM by the beginning of OJ (there are a few examples of conclusive u = 1/wu/1). Thus already in OJ a migration of one-mora UB verbs to UM was under way; this was completed in EMJ. UM comprises only the following simple verbs (in addition there are a few compounds in -mi- 'see', -wi- 'bring along') 12

(98)

i- 'cast (metal)' i- 'shoot' ki- 'put on' mi- 'see, look' mwi- 'tum' (originally UB) ni- 'resemble' ni- 'boil' wi- 'lead, bring along' wi- 'sit down, settle down' (originally UB)

Additional UM verbs from EMJ i- 'pour' [no OJ attestation] moti-wi- 'use' [no OJ attestation] pi- 'dry out' [UB in OJ: pwi-] pi- 'sneeze' [UB in OJ: pwi-] pi- 'winnow' [UB in OJ: pwi-] UM is usually not considered an irregular class, mainly because of its position in later stages of Japanese when eventually all-i- base verbs became UM, but 12

UB arabwi- 'be wild' is in a few instances found with the adnominal arabwiru of the UM paradigm (SM 62, EN 12) as opposed to the regular UB adnominal arahuru (e.g. EN27); EMJ UB isali· 'weep' is in the Kojiki given asisatiru. Both maybe taken as very early, idiosyncratic instances, which did not gain currency, of the later general shift ofUB verbs to UM (see 15.1.3). Arabwi- is ara- 'wild'+ -bwi- and may have shifted for some speakers to UM in analogy with the shift ofUB wi- 'sit down' to UM.

107

3.4 Conjugation classes and morphophonology

it is irregular vis-a-vis the regular vowel bases in the selection and attachment of inflectional endings and in the syntactic use of conjugational forms. There is no attestation in OJ of a UM verb in simple conclusive function. Furthermore, as opposed to other vowel base verbs, this class uses the base as the stem for all formations, including the adnominal and exclamatory forms, and the modal verb extensions be-, rasi-, ram- and the conjunctional particle tomo, which otherwise regularly attach to the conclusive, attach to the base of UM verbs, mi-be-, mi-ram-, ni-rasi-, mi-tomo. All this has been taken to mean that at an earlier stage the base of UM verbs was used as the conclusive form, but use of the stem in conclusive function is not attested in OJ either. From EMJ onwards, UM verbs use the adnominal (miru) also as conclusive. We include the stem in brackets as conclusive in the paradigms for this class to reflect that this form was used with extensions, but not (attested) in conclusive function. In the imperative, -yo is generally used, but mi- 'see' occasionally uses the base as imperative.

mi-

ni- 'boil'

mwi- 'turn'

Infinitive Imperative

mi mi (+yo)

ni ni +yo

mwi mwi +yo

Conclusive

(mi)

(ni)

(mwi)

Adnominal Exclamatory

mi + ru mi + re

ni + ru ni + re

mwi + ru mwi + re

(99)

3.4.3.2 k-irregular and s-irregular These two classes are defined by and consist of a single verb each, ko- 'come' and se- 'do'. (100)

Base

ko-

se-

Infinitive Imperative

ki<=ko+i ko

si <= se + i se (+yo)

Conclusive

ku<=ko+u

su <= se + u

Adnominal Exclamatory

kuru <= ku + ru kure <= ku + re

suru <= su + ru sure <= su + re

From EMJ, two additional verbs which were originally QD and for which some fluctuation was found throughout MJ migrated to s-irr: imase- (OJ QD imas-) 'be, come, go.RESP; respect auxiliary verb', opase- (no OJ attestation) 'be, come, go.RESP; respect auxiliary verb'. Ko- and se- are very similar in conjugation. They are like other vowel base verbs in using the base as imperative and, as with the bigrade and n-irr verbs, their adnominal and exclamatory

108

3 OJ grammar

are formed on the conclusive. They are set off from other vowel stem verbs by having a morphological infinitive which is distinct from the base: ko + i => ki, se + i => si. However, in the prohibitive form, se- uses the base in the prohibitive: na-se-so 'don't do', and in some contexts where other verbs use the infinitive as a stem, they use the base, see 3.4.4.1. 3.4.3.2.1 Grammatical uses of ko- and se- Se- 'do' is not mainly used as a lexical verb. As in NJ, it functions as a pro-verb, standing in for other verbs, and following a verb infinitive se- functions as a supporting light verb, carrying finite predicational markers, as in (101) where the infinitive omopi 'think of with deep emotion' is focused by so (cf. 3.7.2) and se- completes the predication. This function of forming extended predicate structures is shared with ar-, which is used in a similar way with the regular and adjectival copulas and with the negative auxiliary (see 3.4.2.1.1).

(101)

apa-nu pi mane-mi omopi so a ga meet-NEG.ADN day many-ACOP.INF think.INF FOC I GEN suru

do.ADN 'not having been able to see you for many days, I do think deeply and fondly of you' (MYS 19.4198)

Ko- is used as an important and frequent auxiliary verb 'come to ... '(3.1.2). Etymologically, it is likely that a number of grammatical morphemes are cognate with (the sources of) these two verbs, see 3.5.2. 3.4.3.3 -KoseThe auxiliary verb -kose- is an irregularly and incompletely inflected vowel base: kose-. It was already at the OJ stage infrequent and limited in use, and in EMJ went out of use. The meaning of -kose- itself seems to have been 'do for me', but it is only found in fixed optative expressions, either in combination with modals or in the imperative, meaning 'please let there, I wish there'd (/you'd), won't you (please)'. It has the forms in (102). The conclusive is only used as in (1 03) followed by prohibitive na yume (prohibitive particle na and emphatic adverb yume); the base is used as a combinatory stem to form the optative (uti-yame-kosene 'I wish they'd stop') and with the negative -kose-nu (always followed by kama, as in e.g. arikosenu kama 'couldn't there be?; I wish there were'). What is interesting about this verb and what sets it off from a rare and incompletely inflected LB verb is its imperative -koso which seems to use the etymological root in that function. In EMJ a regular, analogically formed, imperative -kose is found in a few examples. Furthermore, the root of this verb may be directly reflected in the focus particle koso.

3.4 Conjugation classes and morphophonology (102)

Base

-kose-

Conclusive Imperative Optative

kosu koso kosene

Negative

kose-n-

109

(1 03) a. ware yukite kapyeri-kuru made tiri-kosu I go.GER return-come.ADN until fall-do.for.me.CONCL nayume PROHIBITIVE 'I wish they wouldn't fall until I have gone and come back' (MYS 15.3702)

b. ume

ga pana tira-zu ari-koso plum GEN blossom fall-NEG.INF exist-do.for.me.IMP 'I wish the plum blossom would remain without scattering' (MYS 5.845)

3.4.4

Extended inflectional forms; combinatory stems

In addition to the five basic inflected forms, the formation of almost all other inflectional forms can be described in terms of combination of a verb stem and an inflectional suffix (comprising both fiectives and auxiliaries). Table 3.11 shows combinations of stems and inflectional suffixes, combining the forms in 3.1.3 and 3.1.4 above, sorted into three sets by verb stem. The majority of inflectional forms are formed directly on the base of regular vowel base verbs, but have either -i- or -a- after the base in the consonant base classes. That is to say, they are either built on the infinitive (3.4.4.1) or on a derived stem in -a (3.4.4.2). Two forms are built on the exclamatory (3.4.4.3). From EMJ, a num her offorms were built on a new stem, the onbin stem (8.1.4). 3.4.4.1 The infinitive as stem The forms under (a) in Table 3.11 are readily segmentable into infinitive plus an invariant (consonant initial) formant: -te, -tutu~ -te-, -n-, -ki, -/eyer-. This is shown most clearly by those verbs which have a morphological infinitive distinct from the base: the consonant base verbs and s-irr and k-irr. The only exception is the use of the base of s-irr and k-irr verbs with the s- initial forms of the simple past: se-si (adnominal), se-sika (exclamatory), ko-si, ko-sika. Ko- does not combine with the k- initial forms of the simple past, but se- has si-ki (conclusive). The suffixes which attach to the infinitive will be noted as in (104). This is a morphophonemic notation to show that these suffixes select the infinitive: /i/ is not part of the phonemic shape of the suffixes and

Table 3.11 OJ forms ordered according to formation on katsuyokei

QD

r-irr

n-irr

LB

UB

UM

s-irr

k-irr

Base

kale-

ar-

sin-

ake-

okwi-

mi-

se-

ko-

(a) Infinitive/base

Gerund Continuative

kaki kakite kakitutu

ari arite aritutu

sini sinite sinitutu

ake akete aketutu

okwi okwite okwitutu

mi mite mitutu

si site situtu

lei kite leitutu

Pelfective Pelfective Simple Past Modal Past

kakite(sakin-) kakiki kakikyer-

aritearinariki arikyer-

siniki sinikyer-

akete(token-) akeki akekyer-

(yodite-) okwinokwiki okwikyer-

miteminmiki mikyer-

sitesinsiki/sesi sikyer-

kitekin-/kosi lei/eyer-

(b) a- stemlb ase Neg.Conj. Optative Conditional

kakakakazi kakana kakaba

arazi arona arab a

sinasinazi sinana sinaba

akeakezi akena akeba

okwiokwizi okwina okwiba

mimizi mina miba

sesezi sena seba

kokozi kozi koba

Respect Causative Passive Passive Negative Negative Conjectural Subjunctive

kakaskakasimekakayekakarekakankakazu kakamkakamasi

akesime-

okwisime-

m)'l?smisimemiye-

sessesime-

kosime-

aranarazu arwnarwnasi

sinassinasimesinayesinaresinansinazu sinamsinamasi

akenakezu akemakemasi

okwinokwizu okwimokwimasi

minmizu mimmimasi

sensezu semsemasi

Ironkozu /romkomasi

(c) Exclamatory Provisional Concessive

kake kakeba kakedo

are areba aredo

sinure sinureba sinuredo

akure akureba akuredo

okure okureba okuredo

mire mireba miredo

sure sureba suredo

kure kureba kuredo

ara-

arasimearaye-

arare-

3.4 Conjugation classes and morphophonology

Ill

will not be included when we give inflectional paradigms of the auxiliaries, as for example in 3.1.4.2 above, and we will segment saki-nu: 'bloom-PERF. CONCL'.

(104)

Gerund Continuative

-(i)te -(i)tutu

Perfective Perfective Simple Past Modal Past

-(i)te-(i)n-(i)ki -(i)kyer-

An alternative analysis, however, does propose that the basic phonemic shapes of these formants have initial -i- (e.g. -ite or -ikyer-) and attach directly to the base. However, this would require otherwise unmotivated morphophonemic rules to derive the correct surface forms, for the regular phonological rules of vowel deletion (cf. 2.6.1) would give incorrect surface forms for the s-irr, k-irr, and polysyllabic LB verbs, see (105b), whereas (105a) gives the correct surface forms:

(105) LB CVCe LBCe S-liT

k-irr

a. Infinitive + -te

b. Base + -ite

ake + te => akete pe + te => pete si + te => site ki + te => kite

ake + ite => *akite pe + ite =>pete se + ite => *sete ko + ite => *kote

It is not surprising that the infinitive is a stem upon which a number of forms are built, for the infinitive is also used as the first member of compounds and may be thought a general compositional form in addition to its independent use, and it is also a nominalized form (the substantive). Thus, the infinitive functions both as a word form and as a combinatory stem. 3.4.4.2 The a- stem The forms under (b) in Table 3.11 are formed by attaching an invariant formant (-zi, -na, -ba; -s-, -sime-, -ye-, -re-, -n-, -zu, -m-, -masi) to the base of the vowel base verbs, but to a stem in -a- for the consonant base verbs. The only exception is the combination of UM verb and respect auxiliary which is lexicalized with those verbs on which it is formed, see 3.1.4.3. The stem of the consonant base verbs used in these combinations is a secondary, bound stem which is derived by adding -a- to the basic stem. We refer to it as the 'a- stem'; it corresponds to the mizenkei in Japanese traditional grammar, see 3.4.6. The suffixes which select the base of vowel base verbs and the a- stem of consonant base verbs will be noted as in (106) to show this. As

112

3 OJ grammar

above, this is an abstract morphophonemic notation: Ia-! is not part of the phonemic shape of the suffixes, and we segment: saka-nu 'bloom-NEG.ADN'. Note, however, that the /a/ of the a- stem diachronically reflects resegmentation of suffixes in initial *a- and that in some cases, /a/ formed part of an earlier shape of the suffix, for example the negative which goes back to *an-; see 3.5.2. (106)

Negative conjectural Optative Conditional

-(a)zi -(a)na -(a)ba

Respect Causative Passive Passive Negative Negative Conjectural Subjunctive

-(a)s-(a)sime-(a)ye-(a)re-(a)n-(a)zu -(a)m-(a)masi

Again, however, an alternative analysis proposes that these suffixes synchronically have initial -a- (e.g. -aba, -asime-) and attach directly to the verb base. This, too, would require otherwise unmotivated morphophonemic rules to derive the correct surface forms, as the regular phonological rules (cf. 2.6.1) would give incorrect forms with the polysyllabic vowel base verbs, (107b), whereas (107a) gives the correct forms. (1 07)

a.

Bigrade base + -ba

UB LB

okwi + ba => okwiba ake + ba => akeba

b.

Bigrade base+ -aba

okwi + aba => **okwaba/*okaba ake + aba => *akaba

3.4.4. 3 The exclamatory as stem Two forms, the provisional and the concessive, are formed by attaching the ftectives -ba and -do to the exclamatory. This is quite clear if considering also the exclamatory, provisional and concessive of the simple past, as in (108). The two suffixes are noted in (109) to show that they select the exclamatory, but as before, /e/ is not part of the phonemic shape of the suffix.

QD

(108) Exclamatory Provisional Concessive

kake kakeba kakedo

LB akure akureba akuredo

Simple past -sika -sikaba -sikado

3.4 Conjugation classes and morphophonology

(1 09)

Provisional Concessive

113

-(e)ba -(e)do

3. 4. 4. 4 The stative and the nominal The stative may be analysed as having the shape -yer- attached to the base of QD verbs: sak- + -yer- => sakyer-; mot- + -yer- => *motyer- => moter-. Diachronically the stative reflects contraction of a construction with the infinitive followed by the existential verb ar-: *saki-ar- 'bloom-is; is blooming'> sakyer-. With s-irr, k-irr and the few UM verbs on which it is formed, a fairly high degree of lexicalization must be envisaged, reflecting the etymological source: *si-ar- > ser-, *ki-ar- > kyer-.

(110) Base Infinitive Stative

QD saksaki sakyer-

QD motmoti moter-

S-liT

sesi seri

k-irr koki kyer-

UM kiki kyer-

The nominal form, too, is thought to reflect an originally analytic construction, consisting of the adnominal form and a following (formal) noun/nominalizer *aku 'thing, place' which is not independently attested. At some point *aku must have been reinterpreted as a flective and the combination morphologized. -kyeku, which is the shape of the adjectival copula as well as a very rare simple past variant, must be thought to be an older lexicalized form, which does not reflect regular synchronic vowel deletion, but diachronic contraction (*ia > ye) at a stage where the adnominal was juxtaposed with *aku. (111)

QD r-rrr -(a)n- Negative LB UB UM k-irr S-liT

n-irr (perfective) Simple past Adjectival copula 3.4.5

kakaku araku -naku akuraku okuraku miraku kuraku suraku -nuraku -siku ~ -kyeku kyeku

<= kaku-aku <= aru-aku <= -nu-aku <= akuru-aku <= okuru-aku <= miru-aku <= kuru-aku <= suru-aku <= -nuru-aku <= -si-aku < *-ki-aku < *ki-aku

Consonant versus vowel base morphophonology

The overall morphophonological difference between consonant and vowel base verbs lies in the extensive use of the base of vowel base verbs. Whereas

114

3 OJ grammar

Table 3.12 Katsuyokei paradigms for OJ verbs

mizenkei ren'yokei shiishikei rentaikei izenkei meireikei

QD

r-irr

UM

n-irr

LB

UB

s-irr

k-irr

kaka kaki kaku kaku kake

ara ari ari aru are are

mi mi miru miru mire mi(yo)

sina sini sinu sinuru sinure sine

ake ake aku akuru akure ake(yo)

okwi okwi oku okuru okure okwiyo

se si su suru sure se(yo)

ko lei

kakye

ku kuru kure

ko

the consonant base verbs use a stem-forming suffix to form the a- stem and ftectives to form the infinitive and the imperative, the vowel base verbs simply use the bare base as the combinatory stem used for forming extended forms, as infinitive and, originally, imperative. In OJ some LB verbs use the bare base as imperative, but mostly the imperative use of the base is reinforced by the exclamatory particle yo. UB verbs generally use yo in the imperative. Of the irregular vowel base verbs, k-irr always uses and both UM and s-irr sometimes use the basic stem as imperative. In OJ yo seems to have been in the course of changing from an exclamatory particle reinforcing the imperative use of the base to an inflectional ending marking the imperative, but it is not yet a ftective proper at the OJ stage. This in itself is an important difference between the vowel and consonant base verbs: despite the superficial resemblance between the imperative formants, the synthetically integrated ftective -ye of the consonant base verbs is not directly related to the fairly independent exclamatory particle -yo.

QD

(112) Base a- stem Infinitive Imperative

3.4.6

kakkak.akak.i(-) kak.ye

LB akeakeake(-) ake (yo)

UB okwiokwiokwi(-) okwi (yo)

The katsuyokei system

The description given above is in many respects similar to the presentation and analysis of verb forms in traditional Japanese grammar where verb forms are said to inflect for six basic 'katsuyokei' (fi§".Jt.l%, inflected forms or stems), which function as wordforms and/or stems used for forming extended forms. Table 3.12 gives the katsuyokei of the eight verb classes (note that only one class has distinct shapes for all six kei, viz. n-irr).

3.4 Conjugation classes and morphophonology

115

3.4. 6.1 Names and uses of the six katsuyokei Mizenkei C*f~% 'irrealis') is a stem used for combination with some flectives and auxiliaries. Ren'yokei (lt.Jt.l% 'adverbial form; the form followed by inflected words (.Jt.l ~ yogen)') is the infinitive, also used as a stem with some flectives and auxiliaries. Shiishikei (~.1.1::.% 'concluding form') is the conclusive. Rentaikei (Jff;$:% 'adnominal form; the form followed by uninflected words({;$:~ taigen)') is the adnominal. Izenkei (Et'~% 'realis') is the exclamatory, also used as a stem with some flectives. Meireikei (-fr'J%% 'imperative') is the imperative. Izenkei and mizenkei are named in contrast after the meaning of the forms ending in -ba: realis: sinure-ba 'as, when one dies'; irrealis sina-ba 'if one dies'. 3.4. 6.2 Traditional names for the verb classes The analysis in the katsuyokei system is intimately linked to the kana script (and is therefore not able to segment forms into units smaller than a mora) and the traditional nomenclature for the verb classes is based on the katsuyokei system in combination with the so-called gojilonzu ('fifty sound table') arrangement of the hiragana and katakana letters in ten columns (gyo rr) by five rows (dan ~), see 6.1.4. The names of the verb classes refer to the positions in the kana arrangement of the final kana used in writing the kei. This terminology pays no heed to the ko-otsu syllable distinctions, as it was devised for the classical language of the late Heian period (where those distinctions had merged) and particularly as it was established before the final discovery of the ko-otsu distinctions; also, transparent endings written in a full kana are ignored for the purposes of these names, namely the -ru, -re and -yo of UM, UB, LB, k-irr, s-irr and n-irr. Quadrigrade (yodan [9~) verbs are thus named because they are written with final kana from 'four rows' {yo-dan), i.e. with four different vowels, e.g. theta, ti, tu, te of the forms of tat- 'stand':

(113)

mizenkei ren 'yokei shilshikei rentaikei izenkei meireikei

tata tati tatu tatu tate tate

t-c."lt t-c."t:J t-c. ""? t-c. ""? t-c.-c t-c.-c

As shown in Table 3.12, the izenkei and meireikei were in fact different (although the difference was neutralized after coronal consonants); thus izenkei (exclamatory) kake was distinct from meireikei (imperative) kakye. However, once /Cye/ and /Ce/ had merged at the beginning of the EMJ period (7.3.2.1), izenkei and meireikei became homophonous. Upper monograde (kami ichidan

116

3 OJ grammar

1::-~) verbs are written with kana from 'one row' (ichi-dan), the -i row (cf invariant mi in Table 3.12), which is in the 'upper' (kami) half in the vertical sequence in the columns in the kana tables. Upper bigrade (kami nidan 1::.=~) verbs are written with final kana from 'two rows' (ni-dan), the -u and -i rows (cf oku, (okwi > EMJ oki)), of which the -u row is in the middle and -i in the upper half Lower bigrade (shimo nidan =~) verbs are written with final kana from two rows, the -u and -e rows (cf aku, ake), of which the-e row is in the 'lower' (shimo) half The final four conjugation classes are termed 'irregular'(~~ henkaku). k-irregular (kagyo-henkaku :nrr~~; ka-hen j]~) is written with kana from the 'k-column' (ka-gyo); s-irregular (sagyo-henkaku -tr"1f~~; sa-hen -tr"~) is written with kana from the 's-column' (sa-gyo); r-irregular (ragyO-henkaku 71f~~; ra-hen 7~) verbs are written with kana from the 'r-column' (ra-gyo); n-irregular (nagyo-henkaku TD'~~; na-hen T~) with kana from the 'n-column' (na-gyo).

r

3.4. 6. 3 Discussion; basic paradigms The katsuyokei system is the standard way of thinking or talking about Japanese verb morphology within the Japanese tradition, but it has not been widely adopted, and is sometimes even denounced, in English language descriptions of Classical Japanese. The katsuyokei system was devised to account for the verb forms of Classical Japanese. To some extent it has the same functions as the four principal parts of Latin verbs. If you know your amo 'I love', amare 'to love', amavi 'I have loved', amatum 'loved', or video 'I see', videre 'to see', vidi 'I have seen', visum 'seen', you can generate and identify any form of those verbs, as well as identify their conjugational class. And likewise if you know the six katsuyokei of a Classical Japanese verb. The katsuyokei framework evolved in the work of Japanese philologists through the Edo period and only found its current form early in the nineteenth century. It is therefore possible that the conception of it was inspired in some way by Latin grammar, which was made available in Japan in Jesuit publications as early as the 1580s. If so, however, this influence remains unacknowledged. All expositions of the history of the katsuyokei framework simply describe it as an indigenous development. First of all, five of the katsuyokei correspond to five basic inflected forms: ren 'yokei = infinitive, shilshikei = conclusive, rentaikei = adnominal, izenkei =exclamatory, meireikei =imperative. Second, in its account of the formation, or segmentation, of most other inflected forms, the katsuyokei system shows that some inflected forms (infinitive = ren 'yokei, exclamatory = izenkei) also function as stems for forming other forms, and as we saw above, the segmentation of the katsuyokei analysis - for example saka-ba rather than sak-aba gives a simpler and more regular account of the formation of the forms,

117

3.4 Conjugation classes and morphophonology

Table 3.13 Katsuyokei paradigms for negative, simple past, adjectival copula and subjunctive

Mizenkei Ren'yokei Shiishikei Rentaikei Izenkei

Negative

Simple past

Adjectival copula

Subjunctive

zu zu -ni zu nu ne

kye- -se

kye leu si lei kyere- kye

mase

lei si silca

masi masi

without recourse to ad hoc rules. Thus, the katsuyokei system captures important facts of the language. It is, on the other hand, also easy to agree that the katsuyokei system has some shortcomings. The major conceptual shortcoming is that it presents the six katsuyokei as equivalent morphological primitives. However, (a) they are not equivalent: some are exclusively word forms (shiishikei = conclusive, rentaikei = adnominal, meireikei =imperative) and some are both word forms and stems (ren 'yokei = infinitive, izenkei = exclamatory); in particular, the mizenkei is not a word form at all, but simply a derived stem, but in applications of the katsuyokei system, the mizenkei is generally treated as a word form on a par with the basic inflected forms. And (b), nor are they morphological primitives. As we saw above, inflected forms, including those presented in the katsuyokei system, are fairly easily segmentable, but it is not possible to capture this in the katsuyokei system which is constrained by the moraic kana writing, precluding a proper analysis of the forms in it. Some presentations of the katsuyokei do provide an analysis of the forms, but it is so bound up with the kana writing that it for example presents the six katsuyokei of kak- as: stem (gokan ~~) ka- + ending (gobi ~fi§) -ka, -ki, -ku, -ku, -ke, -ke, which is of course misleading. Katsuyokei paradigms are also presented for those auxiliaries which do not belong to one of the main verbal conjugation classes (the negative, simple past, adjective and subjunctive auxiliaries), see Table 3.13. The descriptive value of these paradigms is, however, far more limited than for the verbs and they contribute to the confusion about the status of the mizenkei. Finally, some inflected forms, the morphological stative and the nominal form (see 3.4.4.4), are not consistently describable by the katsuyokei system, which also does not include the frequent and important adjectival copula exclamatory- I, -sa, and infinitive-2, -mi. However, bearing in mind its limitations, or perhaps better, its proper application, the katsuyokei presentation is useful in at once capturing basic inflected

118

3 OJ grammar Table 3.14 Basic paradigm for OJ verbs

QD

r-irr

n-irr

Base

leak-

ar-

sin-

a- stem

kaka-

ara

sina

Infinitive Conclusive Adnominal Exclamatory Imperative

kaki kaku kaku

ari ari aru

Ieake

are are

sini sinu sinuru sinure sine

kakye

s-irr

k-ilT

mi-

se-

ko-

mi miru miru mire mi(yo)

si su suru sure se(yo)

lei ku kuru kure ko

LB

UB

LM

Base

ake-

okwi-

Infinitive Conclusive Adnominal Exclamatory Imperative

ake aku akuru akure ake(yo)

okwi oku okuru okure okwiyo

forms and providing a basis for understanding the fonnation of extended inflectional fonns. Especially in the latter function it gives a simple overview of the morphophonological differences between the verbal conjugation classes. We can adjust the presentation a little, as shown in Table 3.14, so that it includes the basic stem, shows that the mizenkei is a derived stem and not a word fonn, and does not include a mizenkei for the vowel base verbs. We will use this mode of presentation and refer to it as the 'basic paradigm' for verbs.

3.5

Proto-Japanese and pre-Old Japanese morphology

3. 5.1

Verb classes; bigrade verbs

There are two facts about bigrade verbs which contribute to an understanding of their pre-history. First, many bigrade verbs take part in transitivity alternations as in (114), where QD (a) and bigrade (b) verbs have opposite transitivity values: (114)

a.

Intransitive tuk- 'stick to' ap- 'meet' yam- 'pause' tum- 'pile up'

b. Transitive tuke- 'attach' ape- 'join' yame- 'stop' tume- 'pile up'

3.5 pJ and pre-OJ morphology Transitive tok- 'untie' war- 'break' yak- 'bum'

119

Intransitive toke- 'be untied' ware- 'be broken' yake- 'be burned'

Second, a number ofbigrade verbs take part in apophonic alternations similar to those exhibited by some nouns (see 2.7.2.2), e.g. (115). In particular, a number of these verbs are clearly related to adjectives. (115)

Bigrade verb base (exposed form)

Adjective stem (covert form)

LB

-e-- -a-

ake- 'redden, lighten'

aka(-) 'red'

UB

-w1-- -u-

sabwi- 'get desolate, fade' opwi- 'get big; grow'

sabu- 'lonely' opo- 'big'

-Wl-- -0-

In these alternations it is the bigrade verb base which corresponds to the 'exposed form', whereas the 'covert form' is reflected in adjective or noun stems or in other verb stems. It is thought that the bigrade verb bases, like the exposed form of the apophonic nouns, phonologically reflect contraction of *Vy diphthongs, so that for example ake- 'to redden' derives from *akay. For the bigrade verbs, the second part of the diphthong is believed to reflect a separate morpheme, thus for example ake- < *akay < *aka-y. We refer to this morpheme *-y as the bigrade 'theme' and to verbs whose basic stem incorporates the theme as 'thematic'. Ohno's groundbreaking study (1953) proposed that the bigrade formant *-y is identical with the OJ QD infinitive formant -i (cf. 3.4.1.3), but since Unger (1977/1993) it has come to be widely accepted that the theme is an originally derivational morpheme, although there is no consensus about its earlier shape or function; the function reconstructed for this formant will have to cover both of the main functional characteristics of bigrade verbs, the derivation from adjectives (115) and the transitivity alternations (114). See Whitman (2008) and Frellesvig (2008) for some recent proposals. 3. 5.1.1 Diachronic classification of verbs The verb classes may be classified diachronically into secondary (thematic) verbs whose basic stem incorporates the bigrade theme and primary (or a-thematic) verbs whose basic stem does not incorporate the theme, see (116). There is no one-to-one correlation between the two major diachronic classes (primary and secondary) and the two major synchronic OJ classes (consonant and vowel base verbs), although all consonant base verbs are athematic; note, however, that it is possible that the n-irr infinitive reflects an older thematic

120

3 OJ grammar

stem: sini ??< *sinu-y or *sini-y. k-irr ko- is primary, directly reflecting its root *b- or *ki. We classify s-irr se- 'to do' as secondary on the hypothesis that it represents a regular contraction from *s:l-y and thus incorporates the bigrade theme~ this is based on forms such as the -so in the prohibitive form (3.1.3.1) which suggests a root so < *s:l which is also reflected in the basic stem: se- < *s:l-y. Other reconstructions divorce the so in the prohibitive from the verb se- 'do' and posit se- < *sye < *se through (partial) mid vowel raising~ if so, s-irr should also be classified as primary. (116)

Primary (athematic) Secondary (thematic)

QD, r-irr, n-irr, UM, k-irr LB, UB, s-irr

I have elsewhere argued that the secondary verb classes constitute a younger morphological layer in the Japanese language than the primary verb classes and that this means that no account should be taken of the forms of the secondary verbs in reconstructing simple pJ verb paradigms. We will not discuss this any further here, but see Frellesvig (2008).

3.5.2

Pre-history ofverb suffixes

Apart from basic inflected forms, most auxiliaries and flectives in OJ attach either to the a-stem or to the infinitive of the consonant base verbs. The a-stem reflects a diachronic reanalysis of the combination of verb stems and suffixes in initial /a/. This reanalysis is shown in (117), exemplified by the combination of *anV, the ancestor of the negative auxiliary -(a)n-, and *yak- 'burn' and *y:ls:l- 'draw close' Y (117) Reanalysis Later (pre-)OJ formation Original pre-OJ formation > {(*yak-=>) *yak-anV => *yakanV} (yakanV <= yak.a-nV (<=yak-)} {(*yasa- =>) *yasa-anV => *yosanV} > (yosan V <= yos. a-nV (<= yos-)}

Note that in the original formation, a base final vowel was deleted in accordance with the vowel deletion rules set out in 2.6.1 above, as in *y:ls:l-anV => *yosanV. There are three parts to the reanalysis: (a) resegmentation of the shape of the suffix from /*-anV-/ to /-nV-/ and conversely (b) resegmentation of the verb stem to which the suffix attached as a derived stem ending in -a- (yaka-, yosa-)~ (c) reinterpretation of the basic stems of all ancestors of QD verbs as having the basic stem shape CVC-, regardless of their etymological root shape (OJ yak-< pre-OJ *yak-, yos- < *y:ls:l-). 13

OJ yos- 'draw close,..' is used as an example of a QD verb originating in an open (vowel final) root; OJ )I?Sor- 'approach' and yose- 'draw closelR' point to a common open root for all three, *y:~s:~.

3.5 pJ and pre-OJ morphology

121

In addition to the negative auxiliary, also the conjectural auxiliary seems to reflect old suffixes in initial /a/, *am V, as both have likely :MK cognates in initial /a/. They are also both involved in the formation of other suffixes. Respect -(a)s- also has a possible :MK cognate, but with a different initial vowel.

(118)

Negative

-(a)n- < *-anV (cf. :MK a·ni negative adverb) -(a)zu < *-ani-su (see 3.1.4.8.3) -(a)zi negative conjectural < *-ani-si

Conjectural

-(a)m- < *-amV (cf. :MK ··amwo 'any, anyone') -(a)ba conditional< *-amu-pa -(a)masi subjunctive< *-amV-(a)si

Respect

-(a)s- < *-VsV (cf. :MK -usi- RESP)

Other suffixes which attach to the a-stem (optative -(a)na (etc.), causative -(a)sime-, passive -(a)ye- and -(a)re-), which have no obvious external links, may also reflect old suffixes in initial /a/, or they may have been analogically formed. In any case, at least some suffixes attaching to the a-stem reflect fairly old Ia/ initial suffixes. The suffixes which attach to the infinitive are different. They are more transparently agglutinating and their use as suffixes seems to be younger. It is possible to view all of them as being derived from a few elements which are also reflected in a number of other grammatical morphemes in OJ. They fall in two groups: (a) forms ink~ s, and (b) forms in t ~ n. 14 It is likely that both sets reflect earlier copulas which were morphologized. The k - s forms involve both of the past tense auxiliaries, -(i)ki and -(i)/eyer- (of which the modal past seems to be derived from the simple past: kyer- < *ki-ar-). The forms of the adjectival copula overlap to a large extent with the past tense auxiliaries, see Table 3.15. This suggests that these are different morphologizations of the same (copula) material, and lends further support to the analysis of the adjectival endings as a restricted copula. Note also that the k ~ s alternation is exhibited by the adjectival copula infinitive -ku and the *-su which takes part in formation of the innovative negative forms (3.1.4.8.3) and which may also form part of the semblative (see immediately below); the morpho-syntactic similarities between the adjectival infinitive and the negative infinitive are easier to explain if, as suggested here, they originate in variant copula forms. It is further possible that the verb se- 'do', the focus particles so and ka (see 3.7.2), and the demonstratives ko and so (see 3.8.3) are root-related to these forms. 14

I have suggested elsewhere (Frellesvig 2008) that the t- - n- forms have Korean cognates.

Table 3.15 OJ adjectival copula and possibly related grammatical forms Adjectival copula fonns Conclusive Adnominal Exclamatory

si

lei kyere < *ki-are

sa

Infinitive Gerund Conditional Concessive Provisional

kyeba < *ki-amu-pa /eye do kyeba

Nominal

kyeku < *ki-aku

Conjectural Negative (nominal)

kyem- < "ki-amkyenaku < "ki-anu-aku

Simple Past Conclusive Adnominal Exclamatory

ki

si sika

ku kute kupa

ku

Modal Past

Negative

Semblative

kyeri kyeru kyere

zu < *ani-su

/eye redo /eye reba

Infinitive Gerund

zu < *ani-su zute < *ani-su-te

Conditional

seba

Concessive Provisional

sikado sikaba

Nominal

siku

Conjectural

nasu < na-su

zupa < *ani-su-pa

kyeba kyeredo kyereba kyelcu kyem-

kyeraku

nasu < na-su

3.6 Verb extensions

123

The t- n fonns are reflected among the verbal suffixes in the perfectives, -(i)te- and -(i)n-, 15 and in gerund -(i)te and continuative -(i)tutu. They are, however, also reflected directly in the simple non-finite copula fonns (3.3), ni, no, to, tu and the particles derived from the copulas. Other related grammatical fonns possibly include nasu which is used after nouns and the conclusive verb fonn to mean 'likeN, as V-ing', e.g. asapi nasu 'like the morning sun'. This construction can be used adverbially or adnominally and in EOJ a variant nosu is found alongside nasu. Nasu- nosu are most likely derived from a copula root, nV-, by means of a fonnant -su which, as mentioned above, may be related to the *-su involved in the refonnation of the negative auxiliary and further to -ku, the infinitive of the adjectival copula. Also the formant -zi which derives denominaljiku adjectives 'likeN' (see 3.2.4.2) may be thought to be -zi < *nVsi < *nV- copula+ -(V)si adjective formant (3.2.4). (119) Copula Case and conjunctional particles Genitive particle Gerund Perfective Continuative Semblative

3.6

t-

n-

to, tu to tu -te -te-tutu

ni, no ni, nite no -nnasu - nosu - zi

Verb extensions

Verb extensions are inflecting clitics. They follow a finite verb form to fonn an extended verb syntagm. They are all modal in meaning: be- necessitive, masizi negative potential, ram- present conjectural, rasi- presumptive, nar- evidential. (120) Pres. conject. Presumptive Necessitive Neg. potential Evidential

15

QD

LB

kaku ramkaku rasi kaku bekaku masizi kaku nar-

aku aku aku aku aku

ramrasi bemasizi nar-

r-rrr

UM

aru ramaru rasi - arasi aru be-

mi rammi rasi mi be-

ari nar-

miru nar-

This is contrary to the traditional account, which diachronically derives -(i)te- from the verb (s)ute- 'discard, throw away' and -{i)n- from in- 'disappear, pass, elapse' and accordingly believes that the function of assertion is a secondary development (e.g. Ohno 1990: 1473). See Frellesvig (2001) about this.

124

3 OJ grammar Table 3.16 Inflected forms of OJ verb extensions be-

masizi-

rasi-

ram-

nar-

Conclusive Adnominal Exclamatory

be-si be-ki

masizi masizi-ki

rasi rasi-ki

ramu ramu rame

nari naru nare

Infinitive-! Infinitive-2

be-ku be-mi

masizi-mi

Be-, rasi-, ram-, masizi-, nar- (evidential) follow the conclusive form of verbs from all other conjugation classes than r-irr and UM (masizi is, however, not attested with UM or r-irr verbs); note that when rasi is used with a r-irr verb, a reduction often takes place: arn-rasi => arasi, -kyern-rasi => -kyerasi. Nar- evidential follows the conclusive of all verb classes (but from EMJ the adnominal of r-irr). The verb extensions exhibit the inflected forms in Table 3.16. Rasi- and masizi- belong to the adjectival shiku (jiku) conjugation; be- belongs to the adjectival ku-conjugation. Ram- belongs to QD and nar- to the r-irr conjugation. Necessitive be-: expresses necessity, obligation and strong probability. This is the morphologically most versatile verb extension, mainly due to its infinitive which forms the basis for extensions with ar-. Negative potential masizi-: is most commonly found with verb forms involving some expression of ability, giving the meaning 'probably/surely cannot'. Masizi- is rare and restricted in use already in OJ; it is not found in EMJ, but is thought to correspond to (to have become) EMJ mazi. Present conjectural ram-: 'probably, apparently, seemingly' usually relates to the present. Presumptive rasi-: 'presumably'. The adnominal is very rare (one example in the Man 'yoshil); instead the conclusive was used in noun modifying function. Evidential nar-: has two main functions: (a) 'it sounds as if somebody does, one can hear somebody do'; (b) 'it is said/reported that somebody does'. (a) was somewhat more frequent than (b) in OJ. Note that evidential nar- is distinct from the assertive use of the homophonous EMJ copula nar-. The assertive copula follows the adnominal form, unlike evidential nar-. Evidential nar- may be thought to be a lexicalized contraction from *na 'sound'+ ar- 'exist'.

3. 7

Particles

Particles are bound postpositional grammatical words which attach to a host, minimally a word. We do not know how closely particles attached to their host in OJ, but sound changes taking place in the first half of EMJ show that particles then were phonologically integrated with the host word (8. 7). Particles contribute to the syntax, semantics and/or pragmatics of an utterance. As the best-studied OJ texts are poetry it is difficult to gain a complete picture

3.7 Particles

125

of the use and functions of the particles. The literary or rhetorical style employed in the OJ poetry means that it is full of exclamations, invocations, lamentations, etc. These are, of course, features of language use in any culture at any time, but this feature of the available OJ materials seems to have skewed the view of the grammar of OJ particles, many of which traditionally are glossed 'emphatic'. In Japanese school grammar, Joshi (.W-J~iil) 'auxiliary word' includes particles, but also a number of verbal inflectional endings, for example the conditional formant -(a)ba. This is because Japanese part of speech classification traditionally is a classification into morpheme types, not into word types. However, the two are also different morpheme types: inflectional endings take part in forming a word whereas particles attach to a full word, phrase, or clause. Note, however, that there is some functional overlap between inflectional verbal endings on the one hand and conjunctional, final, and interjectional particles on the other, in that both contribute to the expression of modality and interclausal syntax. But note also that not all such inflectional endings, e.g. the imperative formant -yo used with vowel base verbs, are included amongjoshi in the traditional classification. The following classification of particles into six types is traditional (although other classifications are found). It is based first of all on functional criteria and several particles belong in more than one class. (121) a. case particles (kaku-joshi f'!.W-J~ii]; 3.7.1) b. topic and focus particles (kakari-joshi 'f*.W-J~ii]; 3. 7.2) c. restrictive particles ifuku-joshi jilj.W-J~jjj; 3. 7.3) d. conjunctional particles (setsuzoku-joshi tiiWE.W-J~ii]; 3. 7.4) e. final particles (shil-joshi ~.W-J~ii]; 3.7.5) f. interjectional particles (kanto-joshi r~,rit.W-J~ii]; 3.7.6) Limiting the classification to particles proper (i.e. excluding inflectional endings) and with the addition of a class not provided in the school grammar, namely that of complementizer (3.7.7), this grouping is by and large valid for OJ and the following stages of Japanese.

3. 7.1

Case particles

Case particles attach to nouns and nominalized forms of verbs and adjectives, specifying grammatical relations within a clause. Although we speak of these OJ particles as case particles, they do not form a fully developed case system comparable to that ofNJ- or of languages with case inflection- nor do they reflect an inherited case system (see 3.7.8). Case marking of core arguments, subject and object will be discussed below in 3.7.1.2. Using familiar names for cases, the OJ case particles are as follows, divided into three groups on the basis of their use and productivity:

126

3 OJ grammar

(122)

Main Accusative Genitive Dative Ablative Comitative

ywori (~ ywo to

Obsolete Nominative Genitive

tu; na

Emerging Ablative Allative

kara pye

WO

no;ga ni

(~

~ yuri ~ yu)

da)

3. 7.1.1 Main Old Japanese case particles Accusative wo is mainly used to mark direct and traversal objects, as in NJ, but was in OJ used more widely also to mark durational adverbials. It is also used as a conjunctional particle and as an interjectional particle. When followed by the topic particle pa the resulting form is wo-ba. Dative ni is the general oblique case, marking both argument and non-argument oblique nominals. The main uses are indirect object, allative, purposive, 16 agent, instrumental, locative, temporal. A variant nite is used in some of the peripheral functions, especially instrumental, locative, temporal. Ablative ywori- ywo - yuri- yu are used about source of movement, comparison, material, and means: 'from, than, with'. There is no discernible difference in function between the four variants; in EMJ only the shape yori survived. Comitative to is used as coordinative, comitative, and comparative: 'with, and, than'.

3.7.1.1.1 Genitives Both no and ga were productive genitive particles. In addition to being a genitive, no was in OJ, and still is in NJ, an adnominal form of the copula (3.3, 15.2). The genitive function of no seems to derive in pre-OJ from the function as adnominal copula (see 3.7.8.2.1), but at the OJ stage both functions of no were firmly established and fully independent. From the common genitive use, no and ga developed differently and have quite different functions in NJ. The major changes are as follows: Ga acquired an additional use as a conjunctional particle ('and, but, as') in EMJ (8.7.2), and 16

In this function ni is also used in purpose-of-motion constructions after verb infinitives, e.g. tumi ni leu 'come to pick (flowers)'.

3.7 Particles

127

changed from a genitive to a nominative case particle in late LMJ (see 12.6.2).

No remains a copula and a genitive case particle, but in early NJ acquired the additional function of a nominalizer (see 12.6.1.3); however, in between, in EMJ and early LMJ, no had more nominative-like functions than ga (see 12.6.2). The different developments of no andgawill be summarized in 12.6.2, but (123) shows the main OJ and NJ functions; it is clear that ga functionally has changed more than no, but it must be emphasized that of the five main functions exhibited by ga and no in NJ shown in (123), genitive, nominalizer, and nominative are in a number of dialects distributed differently over the two particles; in some dialects, for example, no is nominative, whereas ga is genitive and nominalizer. (123)

OJ

NJ

no

Copula Genitive

Copula Genitive Nominalizer

ga

Genitive

Nominative Conjunctional particle

In OJ, genitive no and ga had two main functions (124). Both ga and no were used in both functions. (124) a. Adnominalization (NP-ga/no NP) b. Subject marking (NP-galno VP) Adnominalization is the primary function associated with the genitive in Japanese, as seen in many of the examples cited in other contexts throughout the book. Furthermore, bothga and no were used to mark subjects from OJ onwards. In OJ, this was limited to marking subjects of adnominal clauses and other subordinate clauses, and more rarely non-declarative main clauses (cf. 8. 9). (125) a. kimi no imasi-seba my lord GEN exist.RESP-SPST.PROV 'If my lord had still been here' (MYS 3.454)

b. sawarabi

no moye-duru paru fern GEN sprout-emerge.ADN spring 'In the spring when the fern sprouts' (MYS 8.1418)

c. kimi

ga imasaba my lord GEN exist.RESP.COND 'If my lord is (still there)' (MYS 19.4280)

ni DAT

128

3 OJ grammar

Genitive subject marking is sometimes thought to have originated in reanalysis of relative constructions like (126a) as (b), but whether or not that is the case, genitive marking of subjects in subordinate clauses was firmly established already in OJ and not limited to adnominal clauses. (126) a. wa ga [yuku miti] I GEN go road 'my road to go, that I'm going' b. [wa ga yuku] miti 'the road I go'

3. 7.1.1.2 Differences between no and ga Both no and ga were used in both of the functions in (124), but there were significant differences in the constituents they could mark. Generally ga was far more restricted in use than no: ga was only used to mark noun phrases referring to humans (or personified animals or things), whereas no could mark all nouns, including those referring to humans. This is also reflected in the fact that personal pronouns take ga (wa-ga 'mine', na-ga 'yours', si-ga 'his', ta-ga 'whose'), whereas demonstratives take no (ko-no 'this', so-no 'that') (cf. 3.8.1). The systematic restrictions on the use of ga with nouns continued in EMJ and early LMJ where ga was increasingly curtailed; see further 12.6.2. Importantly, the restrictions on the use of ga applied both when the genitive was used to adnominalize and when it marked subjects. In both cases ga was severely restricted. This shows that it was properties of the host noun which determined whether ga could be used. There was, however, one context which favoured ga over no, namely marking clauses with the predicate in the adnominal form. In the examples in (127), ga functions as a kind of complementizer to adjoin a clause to a head: In (127a) ga connects a modifying clause to a head noun; also no was used in this way (127b), but in OJ and EMJ less so than ga. This use was restricted to ungapped modifying clauses, see further 12. 6.1.1.1. Only ga, never no was used to adjoin clauses to a predicate, as in (127c-d). In the parallel construction in (127c), ga adjoins a clause to the first gotoku, whereas the clause attaching to the second gotoku does not have ga, showing that the use of this construction was optional in this case, although it seems to have been obligatory when connecting a complement clause to adjectives in the exclamatory form (cf. 3.2.2.3), as in (127d). (127) a. wagimokwo ni mise-mu ga tame ni my.beloved DAT show-CONJ.ADN GA sake COP.INF 'In order to show (them) to my beloved' (MYS 19.4222)

3.7 Particles

129

b. taye-mu no kokoro end-CONJ.ADN NO heart 'The intention to end it (our relationship)' (MYS 12.3071)

ga c. puku kaze no mt-ye-nu blow.ADN wind GEN see-PASS-NEG.ADN GA goto-ku, like-ACOP .INF yuku midu no tomara-nu goto-ku go.ADN water GEN cease-NEG.ADN like-ACOP.INF 'like the blowing wind is not visible, like flowing water does not cease' (MYS 19.4160) d. kogu punabito wo mtru ga tomosi-sa row.ADN boatsman ACC see.ADN GA enviable-ACOP.EXCL 'How enviable it is to see the rowing boatsmen!' (MYS 15.3658)

Finally, it is often claimed that the use of no and ga was subject to social differentiation, for example so that ga is said to have been used with 'sentient nouns whose referent is someone close to the speaker or the person who dominates the narrative viewpoint', whereas no is used with 'exalted or indefinite animate nouns' (see Takeuchi 1999: 159-60 who offers (128) as an illustrative example); cf. also 12.6.2.1 about this. (128)

titi-papa ga tame ni moropito father-mother GEN sake COP.INF all.people no tame ni GEN sake COP.INF 'for the sake of father and mother, and for the sake of all people' (Bussoku 1)

3. 7.1.2 Case marking of subject and object The case marking of subjects and objects in OJ has recently attracted a great deal of interest from scholars working in different theoretical frameworks, but from a neutral descriptive point of view, the case marking of subjects and objects in OJ can be summarized as in (129):

(129) Declarative main clauses Subordinate/non-declarative main clauses

Subject 0 0, ga, no

Object wo,0 wo,0

(130) gives examples of these possibilities. As in EMJ, LMJ and NJ, marking of subject and object was optional. It is first of all an important fact of OJ

130

3 OJ grammar

(holding also for EMJ) that it did not have means of explicitly marking the subject of a declarative main clause and the development of a nominative case particle is one of the few major syntactic changes to have taken place in Japanese. This will be discussed in 12.6.2. As mentioned above (3.7.1.1.2), ga was more restricted than no in the subjects it could mark. Marking of objects overall seems not to have changed greatly since OJ: as today, objects could in OJ be marked by wo or left unmarked. While clear-cut conditions which force or bar wo-marking of objects have yet to be established, it seems clear that adjacency between object and verb tends to disfavour wo-marking of objects, although there are plenty of examples of wo-marked objects adjacent to their verb, such as (130e). However, one clear regularity is that if subject and womarked object co-occur in a clause, the wo-marked object precedes the subject, as in (130f). Finally, it should be noted that wo, like accusative markers in many languages, in addition to marking objects can mark a range of durational, temporal and locative adjuncts, e.g. (130g). (130) a. ume no pana ((} tiri plum GEN blossom 0 scatter.INF 'the plum blossoms scatter ... ' (MYS 5.838) b. kimi ga yuku miti my.lord GEN go.ADN way 'the way my lord goes' (MYS 15.3724)

no ki-naku yamabuki warbler GEN come-cry.ADN kerria 'the kerria for which the warbler comes and cries' (MYS 17.3968)

c. ugupisu

d. ume no pana @ wori break.off.INF plum GEN blossom 0 'breaking off the plum blossoms' (MYS 5.843) ume to no pana wo wori green-willow plum COM GEN blossom ACC break.off.INF 'breaking off the blossoms of the green willow and the plum' (MYS 5.821)

e. awo-yanagi

f. wal'l? wo yami ni ya imo ga I ACC dark DAT Q beloved GEN kwopwitutu aru ramu long.for.CONT exist.ADN PCONJ.ADN 'Would my beloved be longing for me in the dark?' (MYS 15.3669)

3.7 Particles

131

g. akikaze no samu-ki asake WO sanu autumn. wind GEN cold-ACOP. ADN dawn ACC Sanu no woka kwoyu ramu kimi cross.CONCL PCONJ.ADN my.lord GEN hill 'my lord, who will cross the hills of Sanu during dawn when the autumn wind is cold' (MYS 3.361) 3. 7.1.3 Obsolete and peripheral case particles Genitives (attributives): Tu which only survived into EMJ in lexicalized collocations was somewhat fossilized already in OJ, in expressions like nipa-tutori 'garden-GEN-bird; chicken'. It is sometimes termed 'locative genitive' as it is often found after nouns denoting some kind of place; this is, however, not likely to be an original feature of this particle which derives from a copula and which is also used to adnominalize adjectives and other words (see 3.2, 3.3). A genitive marker na is usually included in grammars, based on lexicalized forms such as ma-na-kwo 'eye-na-?child; eye(ball)' and ta-na-soko 'hand-na-bottom; palm of the hand'. 17 Based on a few words such as kedamono 'beast' (ke-da-mono 'hair-da-being'), dais often said to be an obsolete variant of na. Nominative i is rare in the OJ texts - about twenty examples in the entire text corpus -and it went out of use after OJ (except for extensive use in EMJ kunten glosses to Chinese texts, see 9.1.1). Most OJ examples of i are found in the Senmyo which are thought in several respects to preserve archaic language usage (and which also are the OJ text type most heavily influenced by kanbunkundoku). There is no consensus about the main or basic function of OJ i. It has for example been studied in Miller (1989) and Vovin (1997) who arrive at strikingly different conclusions, namely that i is an old accusative marker (Miller) or an active marker in a vestigial active/passive alignment system (Vovin). What is clear is that i is used to mark certain subjects, mainly in subordinate clauses. It is also traditionally said to be used for emphasis, and there are some occurrences where it appears to nominalize. This particle is at best marginal already in OJ, but is of some diachronic interest, as it may be related to the Korean nominative particle i, either as a cognate or as a loan into Japanese. 17

Vovin suggests (1994: 253) that na is an old dual or plural marker, based on the occurrence of na mostly with paired body parts (ma-na 'eyes', la-na 'hands') or uncountable nouns (mi-na 'water'). There seems to be only a single compelling counterexample to Vovin's suggestion: momo na pito (}VSK 11) '100-na-person; 100/many people'. However, another possibility is that the forms in -na in fact reflect pre-OJ root final consonants supported by an epenthetic vowel, so that lana diachronically should be segmented tan-a. This hypothesis is based on the observation that this na almost only occurs with nouns that take part in apophonic alternations and would entail that the root final consonant in some cases yodized and contracted with the preceding vowel (see 2. 7 .2.2) and in others was reflected as n + epenthetic vowel a, so that for example *taC 'hand' showed two developments: (a) *taC > *tay > te, and (b) *taC > *tan> lana.

132

3 OJ grammar

(131)

i 1pye naru liDO obobosi-mi house be.at.ADN beloved NOM be. depressed-ACOP .INF se-mu do-CONJ.CONCL 'my beloved who is at home will be worrying about me' (MYS 12.3161)

(132)

nakamaro i itupari kadam-yeru kokoro be.deceitful-STAT.ADN heart Nakamaro NOM lie.INF wo motite ikusa wo okosi ACC hold.GER army ACC raise.INF 'Nakamaro, having a lying and deceitful heart, raised an army' (SM28)

3. 7.1.4 Emerging case particles Kara, which in later periods replaced ywori to become the general ablative marker, was in OJ coming into use, but was not yet fully grammaticalized. It is thought to derive from a noun 'will, way, extent' which was all but obsolete in OJ~ it is often followed by ni and often expresses reason rather than source of motion. It is also found in the conjunctional particle mono kara and had other semi-grammatical uses as well. The nounpye 'side, direction' was being grammaticalized as an allative case particle pye, but in the OJ period had not yet acquired that status.

3. 7.2

Topic and focus particles

Pragmatic particles single out a nominal constituent for comment or as focus for emphasis or question. They express no particular syntactic relationship between the marked constituent and its predicate and may be divided into two groups, topic and focus particles. The topic particles pa and mo are used with little difference from later stages of Japanese, although OJ pa (> EMJ wa) mainly marks a contrastive topic~ mo marks an emphatic topic, 'also, even'. When combining with accusative wo, pa becomes ba: wo-ba. The focus particles mark the focus of emphasis, namo, so, koso, or question, ya, ka. Nama is found only once in Man 'yoshii, but is relatively frequent in Senmyo~ it has the rarer variant namu which replaces nama in EMJ. So has the variant zo which replaces so in EMJ. The focus particles take part in the important construction kakari-musubi which will be discussed in detail in 8.9.

3. 7.3

Restrictive particles

Restrictive particles form adverbial phrases of extent or degree and include: bakari 'about, approximately', dani 'at least~ even', made(ni) 'untit so much

3.7 Particles

133

that', nomwi 'only, solely, alone~ entirely, fully, completely', sape 'also, further, besides, even, at least', si 'also, even', simo 'even', sura 'even, at least'. Note that made(ni) which in EMJ came to be used as a conjunctional particle already in OJ could follow a nominalized clause.

3. 7.4

Conjunctional particles

Conjunctional particles follow finite verb forms to form subordinate clauses: gane (only inpostposed purposive clauses to volitional main clauses) 'so that'~ gani 'as if'~ mono-wo 'although'~ mono-kara (ni) 'as, while'~ mono-yuwe (ni) 'as, while'~ nape (ni) 'at the same time as, together with'~ ni 'as, when, because'~ to (a) concessive '(even) if, although' (in this use, often followed by mo: tomo~ cf. here the etymologically related concessive formant -do(mo}), (b) purposive 'in order to, (so) that'~ wo 'as, because'~ yuwe (ni) 'because' (also after nouns: 'because of'). Most conjunctional particles follow the adnominal, but gane, gani and concessive to(mo) follow the conclusive~ however, to(mo) follows the infinitive of the adjectival copula (na-ku tomo, 'although there isn't') and the base of UM verbs (mi tomo, 'although he sees'). As mentioned, in Japanese school grammar this class of particle includes a number of verbal inflectional endings, namely those forming non-finite verbal forms such as the gerund formant -te or the concessive formant -do.

3. 7.5

Final particles

Final particles occur in utterance final position, after finite verb forms or predicate nominals, contributing to the specification of the modality to the entire utterance: kamo (a) interrogative, (b) exclamatory (after a negative adnominal functioning as an optative: konu kamo 'won't she come?~ I wish she'd come')~ moga(mo) optative (after nominals and adverbials, including infinitive-} of adjectives)~ miyu evidential ('it seems, appears'), see 3.7.8.2~ na (a) prohibitive ('don't!'), (b) exclamatory.

3. 7. 6

Interjectional particles

Interjectional particles form interjections or invocations: ro (rare in the Central dialect, but frequent in Eastern), we, wo, ya, yo. It is worth noting that the imperative verb endings for the secondary verb classes conspicuously originate in interjectional particles (-yo for Central, -ro for Eastern).

134

3 OJ grammar

3. 7. 7

Complementizer

The complementizer to is used after directly or indirectly quoted utterances, see (133), sometimes with the verb of utterance omitted, and in naming constructions, see (134). The latter use is closely related to and is difficult to distinguish from use of to as a copula infinitive in some constructions, such as in (81) in 3.3.1, and the use of to as a complementizer likely derives from the copula used in such constructions. In Japanese school grammar the complementizer is grouped with the case particle to, but the two have different functions. (133)

pa piro-si to ipedo ametuti world TOP wide-ACOP.CONCL COMP say.CONC 'although you/they say that the world is wide' (MYS 5.892)

(134)

yamato no /ami wo akidusima to Yamato GEN country ACC Akizushima COMP 'call the country of Yamato Akizushima' (NSK 75)

3. 7.8

Etymology

ipu call.CONCL

Suggestions of more or less plausible internal etymologies for the particles abound in dictionaries, but are in many cases not persuasive. However, a good number of particles do have fairly obvious internal etymologies, indicating that they either were in the process of being, or fairly recently had been, gramm aticalized.

3. 7.8.1 Nominal sources The directional case particles have conspicuous nominal sources: the ablatives ywori ~ywo ~yuri ~yu derive from the source of the OJ nounyuri 'after(wards)' < pre-OJ *yori, 18 and the emerging allative and ablative particles are based on the nouns pye 'side, direction' and kara 'will, way, extent', respectively. The restrictive particle bakari is from the noun pakari 'estimate, limit' (from pakar- 'to measure, plan'). The conjunctional particles yuwe (ni), mono-wo, mono-kara (ni}, mono-yuwe (ni) incorporate the nouns yuwe 'reason', mono 'being, thing', kara 'will, way, extent', in most cases optionally adverbialized by the copula infinitive ni. As in later stages of Japanese, conjunctional particles are frequently grammaticalized from full or dependent nouns, usually optionally adverbialized by the copula infinitive ni. 18

In order to account for the four OJ variants, a shorter pre-OJ variant, *yo, must be assumed, reflecting either shortening or that *yori itself originally is bi-morphemic *yo-ri. Mid vowel raising gave *yori > OJ yuri, *yo > OJ ywo. The remaining two OJ forms, ywori and yu, would then be analogical, based on yuri - ywo.

3.7 Particles

135

It is most likely that other particles have nominal sources which we can no longer recover with the same amount of certainty, e.g. made (ni) which is optionally adverbialized by the copula infinitive and already in OJ could follow clauses with the predicate in the adnominal form; in EMJ, it came to function as both terminative case particle and conjunctional particle ('until'). As a nominal source, matwo 'target' may be suggested, possibly from pre-OJ form *mato(C). 3. 7.8.2 Verbal sources The restrictive particle si most likely derives from the ancestor of se- 'to do'. A single particle, the final particle miyu, derives from a finite verb form, namely the conclusive of miye-, the passive of mi- 'to see'. Traditionally miyu is not classified synchronically as a particle, but it should be, as (a) it follows the conclusive of verbs, and (b) it is only found in the form miyu, not in any other inflected forms, in this function. It was presumably grammaticalized from biclausal or bisentential constructions.

3.7.8.2.1 Copula A number of particles derive from copula forms: no, ni, nite, tu, and to all functioned as particles and as forms of a copula at the same time; see 3.3.1. This double function has continued into NJ where no, ni, to, and the new form de still are productive forms of the copula, as well as particles. As mentioned in 3.5.2, the focus particles so and kamay have originated in an earlier copula ink- - s-, also reflected in the adjectival copula. 3.7.8.2.2 Roots of other verbs Wo, which functions as an accusative case particle, a conjunctional particle, and an interjectional particle, is sometimes said to have been grammaticalized from the root underlying the existential verb war-; however, it has been shown (by Kinsui 2006) that war- originates in a lexicalized stative form of the verb wi- 'sit down' (cf. 3.4.2.1) which is a less likely source of an accusative particle. The focus particle koso is most likely from the root of the defective optative auxiliary verb -kose- ( < *k::1s::1-i). Note that -koso is used as the imperative of -kose- (see 3.4.3.3), distinct from its use as a focus particle. 3. 7.8.3 External etymology The external etymology of OJ particles has mainly been studied within an Altaic, or Japanese-Korean, comparative framework. However, particles are short words and the risk of chance resemblances is therefore not small. More importantly, it is a fact which is not usually addressed explicitly in comparative studies that no significant subset of particles within OJ represents an inherited system. This is well illustrated by the case particles of which only ga, i, wo and rwo lack a transparent internal source, as opposed to no, tu, ni, ywori, to,

136

3 OJ grammar Table 3.17 Pronominal fonns of OJ

Personal 1st

2nd 3rd interrogative reflexive

Short

Long

wa, a na

ware, are nare

Locational

si ta ono

tare

ko

kore

Demonstrative

proximal non-proximal interrogative

YO

idJt-

idJtre

koko soko idJtku

na, kara, pye whose grammmaticalization from other parts of speech is either documented or easily recoverable. This suggests that comparison of individual particles with forms in other languages has to be interpreted with caution.

3.8

Pronouns

The main pronominal forms of OJ are shown in Table 3.17. In addition to these forms, there is an interrogative noun nani 'what' which does not form part of any morphological system. There are also a number of alternative terms of address, see 3.8.2.1, and there is a trace of an earlier proximal demonstrative i, see 3.8.5. Some descriptions include the noun woto-~woti 'distant place or time' as a 'distal' demonstrative, but that is a lexical noun.

3. 8.1

Short versus long forms

A conspicuous feature of the pronominal system is the existence of short and long forms. The original function of the-re of the long forms is not known~ it has been hypothesized to reflect the plural marker -ra +case particle -i, *ra-i > -re (Vovin 1997). Whatever the original grammatical distinction between short and long forms, it is not systematically reflected in their use and distribution in OJ. In OJ, the two sets offorms are formally in free variation in many environments, but in complementary distribution with respect to some criteria: (135)

short Use in isolation Use with genitive particles Use in nominal compounds

long

+ + +

3.8 PronoWls

137

The long forms are used in isolation, i.e. without a following particle, as subject/topic and as emphatic and exclamatory forms, e.g. (136a, b). The short forms are not used in this way, but must be followed by a particle. Conversely, the long forms are not used with genitive particles (apart from a small number of examples of idure-no (136b), which has no corresponding independent short form, and of kore-no) nor as a modifying first element in a nominal compound. The overwhelming majority of examples of the short forms are with genitive particles, used both attributively and as subjects in subordinate clauses: wa ga kokoro 'my heart'; wa ga mure-inaba 'when I go away' (KK 4). The personal pronouns (a, wa, na, si, ta, ono) take ga and the demonstratives (ko, so) take no. 19 The contraction involved in forms such as wagipye 'my house' suggests that in some cases short form +genitive particle had been univerbated into a possesive pronoun, here waga 'my', so that wagipye < waga ipye (see 2.6.1). In addition, the short forms (except for si and ta) are used to form compounds, e.g. a-duma 'my wife', wa-dori 'my (own) bird', na-dori 'your bird', ko-yopi 'tonight'; usually the second member of the compound undergoes rendaku (2.6.2). Note that interrogative idu- is found only as a constituent of derived forms and compounds (except for a single example in the meaning 'where' in EOJ, MYS 14.3549), never with a particle, genitive or otherwise. Thus, use of the short forms is quite restricted, most examples being in compounds or with a genitive particle; this was the only use of the short forms which survived productively into EMJ. In OJ they are, however, also found with other case particles and with focus particles, which seems to reflect an earlier wider use as free pronouns, e.g. (136c). There are even parallel examples with full equivalence between short and long forms, such as KK 4 which in a repetitive sequence has ... kore pa pusapazu 'this will not do' (pusap'befit, be suitable') in the first instance, but ... ko mo pusapazu 'this too will not do' in the second. In contrast with the short forms, the long forms are used more freely, but it must also be noted that some long forms were not well established in OJ Non-proximal sore is not attested in any of the core sources and has only two marginal attestations (in the Kakyo hyoshiki and in handwritten material). Thus sore, which became very frequent from EMJ onwards, was not part of the productive OJ system and is not included in Table 3.17 above. Likewise, reflexive onore is only attested twice (MYS 12.3098, 16.3883) and 2nd person nare is textually rare. Both are frequent in EMJ. Also proximal kore is not as highly frequent as in following periods. Thus, rather than presenting a system with a stable grammatical distinction, such as combining (=short) versus free (=long) forms, the OJ pronominal system was in a state 19

There is a small number of examples of ga used with the demonstrative so when it functions as an inanimate 3rd person pronoun, e.g. so ga pa 'its leaves' (KK. 101), which however has so no pana 'its flowers' in the next line.

138

3 OJ grammar

ofmorphologicaltransition with long forms replacing older short forms whose use was gradually being circumscribed, regardless of what the original grammatical distinction between them may have been. (136) a. wegusi ni ware wepi-ni-kyeri get.drunk-PERF-MPST.CONCL sake.of.smiles DAT I 'I have become drunk on the sake of smiles' (KK 49)

b idure no sima ni ipori se-mu ware which COP.ADN island DAT hut do-CONJ.ADN I 'I! on which island shall I make my hut (for the night)' (MYS 15.3593) pa mo yo mye ni si areba I TOP ETOP EMPH woman COP.INF EMPH exist.PROV na wokite [= wo okite] wo pa na-si you ACC.put.GER man TOP exist.not-ACOP.CONCL 'Me, because I am a woman, apart from you I have no man' (KK5)

c. a

3.8.2

Personal pronouns

The personal pronouns form a morphological class by taking ga, never no, as genitive marker. Furthermore, the personal pronouns, as opposed to nonpronominal terms of address, were not used with plural markers in OJ. This may be taken to support Vovin's etymology for the long forms as involving a plural marker, but note that 1st person are seems not to have been used with plural reference. 1st and 2nd person pronouns were used frequently in OJ, much more than in later stages of the language. Among the 1st person forms, the a- forms are only used for exclusive, singular reference ('I, me (alone)'), whereas the wa- forms also are used for inclusive and plural reference ('me/ us (including you)') and also reflexively ('myself, oneself'), suggesting that the wa- forms reflect an earlier indefinite personal pronoun 'one'. The waforms were used more than the a- forms and there is also some overlap in usage, indicating that a shift from a- to wa- as the 1st person pronoun was in the course of being completed; the a- forms went out of use and are not used in EMJ. An EOJ form wanu, corresponding to central ware, is attested in a few cases. Maro, which in EMJ was used as a 1st person pronoun with some frequency, is attested in a song found in both the Kojiki (KK 48) and Nihon shoki (NSK 39) in the phrase maro ga ti 'my father'. The 3rd person si is used both with animate and inanimate reference. Si is not used much, however; 3rd person reference is mostly expressed by the

3.8 PronoWls

139

non-proximal demonstrative so, from which si is diachronically derived (explaining why si alone among the short forms does not form compounds): *si-i > si. Si is sometimes said to be used for 2nd person reference, but the few examples which may be cited in support of this are not persuasive. Ono is sometimes believed to have alternated with an obsolete word ana 'self'~ na is sometimes said originally to have been used for the 1st person and to be a reduced form of ana, but, again, with little good evidence. 3. 8.2.1 Other terms of address In addition to 2nd person pronouns, OJ had a number of terms of address of which the following are the most prominent. Whereas the personal pronouns proper do not combine with plural markers, some of the alternative terms of address do.

(137)

imasi, masi, mimasi 'you, respectful'~ cf. imas- 'be, exist, respect' ~ mimasi is thought to be more respectful than imasi and masi and to be from the honorific prefix mi- + imasi~ another possibility is that mi represents a heavily nasalized initial [i]masi kimi 'you (my lord), respectful' namuti 'you' (originally respectful, but neutral at the OJ stage~ < na 'you'+ -muti 'esteemed person~ honorific suffix in names and titles', cf. mutu- honorific prefix) namutati 'you, plural' (thought to be from namuti-tati) wake 'you, pejorative' (also '1, humble') (<'lowly person') ore 'you, pejorative' i 'you, pejorative' (only i-ga 'you-GEN')

3.8.3 Demonstratives Most accounts of OJ demonstratives posit a three-term 'proximal- mesialdistal' system, built on ko-so- ka. However, there is no evidence within OJ of ka being a productive member of the demonstrative system. Two forms are attested in OJ: long kare is found once, see (138d) below~ what may be taken to be short ka, as distinct from the adverb ka 'this way', is attested at most twice, both in EOJ poems (MYS 14.3565, 20.4384). While these forms most likely represent the budding of the distal demonstrative which is so frequent in EMJ, they did not form a central part of the OJ system of demonstratives. Other ka-based forms often cited are in fact attested only from EMJ. The description of the semantics of the OJ demonstrative system is due to Hashimoto Shiro (1966) whose study is the first to consider the OJ system on its own merits, rather than in terms of the EMJ system. The ko- versus sosystem is entirely speaker based, with no primary reference to the hearer.

140

3 OJ grammar

Proximal, ko-, refers to what is within the speaker's domain of direct sensory perception, or experience. Non-proximal, so-, refers to what is outside of the speaker's domain of direct experience. The facts of the use of the mainko- and so- forms in OJ are as follows:

• The ko- forms are almost entirely used deictically, referring to what may be directly experienced by the speaker~ the only form used anaphorically is koko. • The so- forms are mostly used anaphorically, with some examples of reference to something which is implied, but has not been mentioned explicitly~ temporal deictic reference to past events is not infrequent. This anaphoric, or conceptual, and temporal deictic reference follows from the definition as being outside the speaker's domain of direct experience. There are no clear examples of spatial deictic use of so~ also soko is mostly used anaphorically, but there are a few examples of spatial deictic use with reference to the hearer. The ternary 'proximal - mesial - distal' system of EMJ and later arose, Hashimoto argues, through a subdivision of the direct domain into 'close vs. far', with the form ka being drawn in as an altemant of ko. Note that the reference of the single OJ example of kare, (138d), is within the field of direct visual perception of the speaker. The few examples of soko with spatial deictic use form part of the development of the three-term system. In addition to the three main forms built on the proximal- non-proximal - interrogative/indefinite bases ko - so - i included in Table 3.17, there are a num her of other forms, see Table 3.18. While these forms show the pervasiveness of the ko - so - i system, and no trace of a distal ka, it is also clear that the system is not as well developed as in EMJ and later stages. Some of what later become derivational morphemes expressing syntactic and/or fairly definite semantic categories, were at this stage apparently semantically vaguer. The designations for the derived categories are those used for later stages of the language~ it is not clear that they are entirely appropriate for OJ. Locational -kol-ku is thought to be from an obsolete noun -kal-kol-ku 'place', cf. e.g. miyako 'capital~ palace-place'~ note, however, the widespread use of short ko to mean 'here', as well as the use of koko and soko without locational meaning. Likewise, koti is not particularly directional, but rather meant 'this way, this side'. Recall that sore was not yet a productive part of the system (3.8.1). The secondary -du- in the interrogatives (of unknown meaning and origin) was reinterpreted as part of the base in what emerged as the productive system~ idu later changed to the do of the well known ko - so - a - do system of NJ. Other OJ forms built on i-du-: idura 'where(abouts)', idupye 'which direction'

3.8 PronoWls

141

Table 3.18 OJ demonstraiive forms

Short Long, -re Locational, -kol-ku Directional, -ti Degree, quantity, -kVI-ku Manner Manner, (-kuf)-ka Time, -tu

Proximal

Non-prox.

Interr.

ko(-)

so(-)

i-lidu-

ko kore koko koti

so

iduidure iduku iduti

kokV-

sokV-

ka kaku

sale sika

soko

ikuika itu

(pye 'side'), idusi 'which direction' (-si 'side'). Attested OJ forms built on i-ku- 'how much' -ko-kV- 'this much' -so-kV- 'that much' with the derivational elements -ra, -da, -ba and the adverbializer -ku: ikura, ikuda; kokoda, kok(w) ida, kokodaku, kok(w)idaku, kokoba, kokobaku, kok(w)ibaku; sokoraku, sokidaku, sokoba.

pa wasurezi ko no tatibana wo I TOP forget.NCONJ this GEN orange.blossom ACC 'I will never forget it, this orange-blossom' (MYS 18.4058)

(138) a. are

b. wakare-ko-si

so no p1 part-come-SPST.ADN that GEN day 'that day when I left' (MYS 17.3978)

c. uwe-si

ta mo maki-si patake plant-SPST.ADN paddyfield ETOP SOW-SPST.ADN paddyfield mo asa-goto ni sibomi-kare-yuku. So ETOP morning-each DAT wither-dry-do.more.and.more That wo mireba ... ACC see.PROV 'For each morning the planted rice fields and the sown fields wither and dry out more and more. When I see that ... ' (MYS 18.4122)

d. a gamopu [= ga omopu} kimi ga mi-pune kamo kare I GEN.think.of.ADN my.lord GEN HON-boat Q that 'is it the boat of my beloved lord, that' (MYS 18.4045)

142

3 OJ grammar

3.8.4 Basic pre-Old Japanese pronominal system Based on the OJ forms, a simple and basic pronominal system may be posited for pre-OJ, see (139), with a 'speaker- non-speaker- interrogative' system for both the personal and the demonstrative pronouns, and including wa as an indefinite: (139)

Personal speaker non-speaker indefinite interrogative Demonstrative speaker non-speaker interrogative

a na wa ta

1st 2nd 'who'

ko so

proximal non-proximal 'which'

i-

ono

Refkxive

3. 8. 5 Proto-Japanese demonstratives Although the OJ system of demonstratives clearly is a two-way 'speaker nonspeaker' plus interrogative system, it seems in fact that it goes back to a three-way 'proximal- mesial - distal' plus interrogative system, like that found in EMJ and later stages of the language, and like that found in Korean, as proposed by Frellesvig and Whitman (2008), see (140) and (141). (140) pJ

Proximal *i

Mesial *ki

Distal *si

Interrogative *e

pre-OJ.a

Proximal *i

Mesial *ki

Distal *si

Interrogative *i

Participant *ki

pre-OJ.b

Non-participant *si

OJ

Speaker ko (- i)

Non-speaker

EMJ

Proximal

Mesial

Distal

ko

so

ka

so

Interrogative *i Interrogative i-- iduInterrogative i-- idu-

The system posited for pJ includes interrogative *e and proximal *i, lost by the time of OJ, but vestigially attested in a lexicalization such as ima 'now' < *i + ma 'space'. The change between pJ and pre-OJ.a of interrogative *e > *i is a result of mid vowel raising (see 2.7.2.3), resulting in homonymy

3.8 PronoWls

143

between proximal and interrogative. This homonymy (between two paradigmatically opposed terms) was resolved by eliminating proximal *i and reinterpreting *ki and *si as speech event participant and non-participant, respectively, in pre-0 J. b and then in OJ as speaker and non-speaker. Typological pressure then resulted in the system being augmented with a distal ka, but not until EMJ. It has long been observed that the J\lfK mesial and distal demonstratives, ku and tye, respectively, present a good form fit with Japanese proximal and mesial ko and so, but that the semantics do not fit. The changes between pJ and OJ explain that; and the pJ system is a good phonological and semantic fit with the J\lfK system. Proximal

Mesial

Distal

Interrogative

J\lfK

i

ku

tye

pJ

*i

*ki

*si

e *e

(141)

REFERENCES General: Iwai 1970, Vovin 2005, 2008, Yamada 1913/1954. Verb morphology: Frellesvig 2007. Derivatives: Martin 1987: 790ff., Unger 1977/1993: 127ff. Uses of the adnominal and nominal forms: Wrona 2008. Tense and aspect Takeuchi 1987, 1999: 10lff, Washio 2002. Adjectives: Kawabata 1976, Martin 1968, 1987: 801ff., Unger and Tomita 1983, Yamazaki 1992. Pre-OJ verb morphology: the seminal article is Ohno 1953; further: Frellesvig 2001,2008, Hattori 1979, Martin 1987, Ohno 1978a, Unger 1977/1993, Whitman 2008. Particles: Ga, no, wo and OJ syntax: Kuroda 2007, Shibatani 1990: 333ff, Yanagida and Whitman 2009a, Wrona and Frellesvig 2010. Focus particles: Ohno 1993, Takeuchi 1999: 169ff, Whitman 1997, Wrona 2008: 195ff. Etymologies: Frellesvig 2001, Martin 1990, 1991, Miller 1971, Murayama 1957. PronoWls: Takeuchi 1999: 180ff, Whitman 1999, Yanagida and Whitman 2009b. Demonstratives: Frellesvig and Whitman 2008b: 27ff, Hashimoto 1966, 1982.

4

Loanwords

The etymological study of the lexicon of OJ, the question of its provenance, and the sorting of inherited from borrowed wordstock are part of the issue of the genetic affiliation of Japanese: OJ words which are similar to, or take part in sets of sound correspondences with, words from another language may be cognate with these words; they may be borrowings from the other language; they may themselves be the source of borrowings into the other language; or the similarities may simply be due to chance. We are more fortunate for the later historical stages of Japanese, but for the OJ language as it presents itself to us, we simply very often do not know and many attempts to identify old loanwords within OJ are highly speculative. It is beyond doubt that OJ includes old loanwords from the languages around Japan - especially words relating to agriculture, seafaring, warfare, spiritual and religious life, government, and administration - but that we will not be able to identify many of them as loanwords on other than extralinguistic grounds. It is for example a strong hypothesis that OJ iraka 'roof, roof tile' is a loanword, but we do not know from where. In other cases we believe that a word must be borrowed and can come up with several likely sources but cannot choose between them. It is, for example, very likely that the Japanese word for 'horse', OJ uma, is borrowed and there are indeed words in surrounding languages which mean 'horse' and which are similar to uma, e.g. EMC *mai', :MK ·mol, Mongolian morin. Mongolian and Korean may be genetically related to each other and/or to Japanese, so the :MK and Mongolian forms may be cognate, but one language may also have borrowed the word from the other. And is the Chinese word the source of borrowing, or, perhaps more likely, itself borrowed? On the other hand, note should be taken of the recent work of Unger (2001, 2003) which provides a linguistic basis for identifying candidates for loanwords. Unger brings attention to a num her of cases in which a word in Korean (e.g. :MK pal 'leg, foot') corresponds phonologically to a Japanese word with a narrower meaning (pagi 'shin'), whereas the Japanese word (asi 'leg, foot') which is synonymous with the Korean word is unrelated to it, that is: 144

4.1 Ainu

(1)

OJ pagi 'shin' asi 'leg, foot'

145

Jv1K ·pal 'leg, foot'

Unger argues that in such cases it is likely that the pagi type word is inherited, but was displaced, and became semantically specialized, by the asi type word which may either be from a substratum or a super- or ad-stratum, much like the displacement and specialization in English of Anglo-Saxon tell and loft by count and air borrowed from French. On the other hand, it is of course quite possible that displacement or replacement of the asi type word was entirely internal and not motivated by borrowing. Linguistic borrowing presupposes some degree of bilingualism. It can be difficult to draw a line between loanwords and foreign words. Foreign words are not infrequently used by bilingual speakers. This is known as code mixing and takes place for various reasons including prestige, novelty, or clarity. A foreign word used in this way can gain currency and eventually be accepted by members of a speech community as part of their language. In the transitional phase the word will be foreign to some speakers but native to others. It may be pronounced with the phonology of the foreign language, with an adapted native phonology, or totally phonologically (and morphologically) assimilated to the native language. It is only in the latter situation we shall talk about loanwords as an established part of OJ. This point is particularly relevant when considering loanwords from Chinese. It is traditional to distinguish between 'loanwords' and 'Sino-Japanese' vocabulary, the former usually designating loanwords from other languages than Chinese. Here the distinction will be made between fully assimilated nativized loanwords, including those of Chinese origin, and the Sino-Japanese vocabulary which has been used through the history of Japanese. The two are different and only the former will be discussed here. Sino-Japanese will be discussed in Chapter 9, but it should be said here that Chinese served not only as a source of Chinese vocabulary but also as a medium for the introduction of much Buddhist vocabulary originally from Sanskrit. Sacred names and terms were usually not translated into Chinese, but adapted phonologically and in writing transcribed phonographically (1.1.2. 4).

4.1

Ainu

It is difficult to identify loanwords in OJ from Ainu, perhaps with one notable and remarkable exception: OJ kamwi ~kamu- 'spirit, deity' may well be borrowed from an ancestor of Ainu kamuy 'bear; deity' (although the opposite direction of borrowing has also been proposed). Recently it has also been

146

4 Loanwords in OJ

suggested by Vovin (2009) that some OJ words (including place names) attested only in EOJ (see Chapter 5), e.g. sida 'time', in fact reflect prehistoric borrowings from Ainu. Though not lexical borrowing as such, it should be mentioned that Ainu words are also preserved in some place names in Japan. Best known are those ending in -betu, Ainu pet 'river', or -nai, Ainu nay 'stream, valley, river'. It is likely that many other place names reflect now unrecognizable Ainu words. 4.2

Continental loanwords: Korean, Chinese, Sanskrit

We do not know what languages other than the ancestor of Japanese were spoken in the Japanese archipelago in pre- and proto-historic times. The time depths involved make it hazardous to attempt to identify borrowed (or sub-stratum) vocabulary from Austronesian or Austro-asiatic, although some languages belonging to those families may have been spoken in Japan at the same time as Japanese. For example, OJ has a word tape 'barkcloth~ cloth made from the bark of the mulberry tree' which must somehow be related to the Austronesian word tapa which means exactly the same, but we can say nothing about the direction or route of transmission. However, it is possible to point out a number of OJ words which in all likelihood are loans and for which we can plausibly identify a source, from Korean, Chinese or Sanskrit. In proto-historic times, many Japanese speakers would have had some facility in one or more foreign languages, ranging from fishermen and traders communicating with their continental colleagues to highly learned clergy. Different continental languages, especially Korean languages and varieties of Chinese, would have been used at different times and particularly within limited social or professional circles. Most continental culture was in pre- and proto-historic times transmitted to Japan via the Korean peninsula. Today Korean is one language with dialects which all descend from Middle Korean, which in turn continues the linguistic tradition of the Kingdom of Shilla. Until unification under Shilla in 668 there were, however, three main kingdoms on the Korean peninsula: Koguryo, Paekche, and Shilla. From the early fifth century contact with and immigrants from the kingdom of Paekche seem to have played a particularly significant part in the transmission of continental material and intellectual culture, including Chinese language, writing, learning, and also later, Buddhism which seems to have been introduced from around the middle of the sixth century. With this came new vocabulary. In addition to loanwords from Korean languages, much borrowed vocabulary of ultimately further origins must be thought to have entered Japanese through some Korean language, or through Korean speakers, but in most cases details remain obscure.

4.2 Continental loanwords: Korean, Chinese, Sanskrit

4.2.1

147

Korean

The following words have been thought to have been borrowed from a Korean language. Usually we do not know which, but must simply assume an earlier cognate form of an attested Middle Korean word as the source, (2a) below, but in a few cases we can tentatively identify Paekche as the source language (reconstruction of Paekche forms follows Bentley 2000), see (2b). If Japanese and Korean are genetically related, some of these words may in fact be cognate rather than borrowings, for example kudira, para, or uri. (2) a.

b.

4.2.2

karamusi 'ramie (fabric, cloth), Chinese silk plant', MK mwosi 'ramie fabric, cloth'; cf. OJ kara- 'China, Korea, foreign' kasa 'bamboo hat, umbrella', MK kas 'id.' kudira 'whale', MK kworoy 'id.' mori 'woods', MK ·mwoyh 'mountain', pK *mwo"lih para 'field, plain', MK pel 'id.' patake 'field', MK path 'id.' pyera 'spatula, pallet', MKpyet 'moldboard' sarapi 'spade' (EMJ), MK salp 'id.', possibly further from OC *tshrap; cf. OJ sapi, sapye 'spade' (see (3a) below) sitogi (EMJ) 'rice cake for ceremonial purposes', MK "stek 'rice cake' uri 'melon', MK ··woy 'cucumber', pK *wo·li kopori 'district', Paekche *bp;)ri, MK kwo wolh kuti 'hawk', Paekche *kuti 'falcon' kwi 'fortress, walled city', Paekche *ki 'id.' sasi 'walled city', Paekche *casl, MK ·cas 'id.'

Chinese

The following may be thought to be early loans from Chinese. At least some of the words in (3a) are quite old and represent direct borrowings. Those in (3b) are transparent and probably not very old borrowings from Chinese, but they are not usually thought of as SJ vocabulary; two (gakwi and zeni) have initial media, which were not allowed in the native vocabulary (2.7.1.2). In Chinese historical phonology EMC and LMC are fairly securely reconstructed and more recently our understanding of OC phonology, too, is improving. We are therefore sometimes able to determine which stage of Chinese a borrowing originates in. (Reconstructions ofEMC and LMC are from Pulleyblank (1991 ); the reconstructed OC forms follow Miyake (1997 and p.c.).)

148

4 Loanwords in OJ

(3) a.

kama 'pot', ±tt OC *khaam kama 'sickle',. OC *gryam ke 'spirit' %. EMC *khil, OC *khiys kinu 'silk', *~ OC *kwyans (EMC *kjwianh) kuni 'country', f~ OC *guns (EMC *gunh) saga 'characteristic; good omen', OC *saiJ (EMC *siajiJh) t-~ EMC *ziaiJ sapi, sapye 'spade',~ EMC *t~h;)ip/t{>s:p, OC *tshrap. Cf. sarapi (see (2a) above). It is possible that OJ sapi/sapye was borrowed from EMC while the ancestor of J\1K salp was borrowed from OC, further being borrowed into EMJ in the shape sarapi sugu-roku no saye 'pair-six-Gen-game' (name of a game of dice), written .ll{:t\]J{6::~ inMYS 16.3827. Sugu is the traditional reading of .ll{ OC *sroiJ, EMC*~aiWIJ/~iWIJ; saye which is spelt out phonographically reflects WLMC *saj' 'game (of dice)'. From EMJ this word became sai ume 'plum,, oc *hmay

·r!

m

b.

4.2.3

gakwi M!~ EMC *IJah kuj' 'glutton, hungry ghost' pakase 1W± EMC *pak ~i' 'expert, authority'. This word is not phonographically attested in OJ, but was surely used as it was an important official title in the ritsuryo system. The regular SJ character readings were used in coining the SJ word cNJ hakushi '(academic) doctor, PhD' puse :tp~ EMC *p:>h c;:ih 'temple offering, charity' (Chinese loan translation of Skt. dana 'offering, alms') saka R EMC *tc;:iajk 'unit of measure; shaku' we ii!U EMC *)'Wail 'picture, drawing' zeni (EMJ) ~ EMC *dzian 'money'

Sanskrit

The words in ( 4) below can be traced back to Sanskrit (or, in one case, Pali). Almost all ultimately derive from Buddhist contexts but became everyday words and most remain in use today. A few words are included which are not attested until EMJ but which were probably in use in OJ. In a few famous examples we can trace the route of transmission and find both Chinese and Korean related forms ( 4b), but for several of the oldest and most naturalized loanwords that is not possible ( 4a). It is worth noting that in addition to Buddhist inspired vocabulary, a word for 'rice', which surely must be old, is

4.2 Continental loanwords: Korean, Chinese, Sanskrit

149

of Sanskrit origin. It is also worth noting that most of these words survive into the modern language. (4) a.

ama (EMJ) 'nun', Pali ammii 'mother' kapara (EMJ) 'ceramic roof tile', Skt. kapiila 'cup, jar, dish; cover' kasa 'scab, the pox', Skt. khasa 'itch, scab' mara 'penis', Skt. mara 'death; the evil one, the tempter; god/passion of love' pata 'banner, standard', Skt. patiikii 'id.' sara 'plate', Skt. sariiva 'shallow cup, dish, plate' uru- (EMJ) in e.g. urusine 'nonglutinous rice', Skt. vrfhi 'rice' (OJ -sine (- ine-ina-) 'riceplant')

b.

potoke 'Buddha, Buddha image', 11Kpwuthye 'Buddha', {.b.WE EMC *but tba ¥¥~ OC *buu daa, Skt. buddha 'Buddha' pati 'bowl', J\1K pali, ~ OC *pat, Skt. piitra 'vessel' tera 'temple', J\1K tyel 'temple', *U EMC *t~bait < OC *tshraatlk.sraat, Skt. k~tra 'place' kyesa 'priest's robe'~~, EMC *kai ~ai; LMC *kja: ~a:, Skt. ka.siiya 'the (yellow) robe of Buddhist clergy'. In addition to preserving kesa in the original meaning, NJ also reflects this word in oogesa-na 'pompous' (oo- 'big').

The words in (5) are some examples from the EMJ period of common Buddhist names and terms ultimately deriving from Sanskrit, but which were taken in via Chinese renditions. (5)

aka 'Buddhist water offering; container for this; wine (priests' secret language)' !Ml11JD, EMC *?at gia, Skt. argha, 'value', arghya 'water offered to a guest at a respectful reception' amida 'Amitabha Buddha' ~iiJ5~WE, EMC *?a mji tba, Skt. amitiibha bosatu 'Boddhisatva' ~~, EMC *b:> sat, Skt. bodhisattva butu 'Buddha, Buddhism' {.b.(~t), EMC *but (tba), Skt. buddha daruma 'Bodhidharma' Jtm, EMC *dat rna, Skt. bodhidharma naraku 'hell' ~1{, EMC *nab lak, Skt. naraka 'id.' setuna 'instant' *Ut.!~, EMC *t~bait nab, Skt. k~atJa 'instant, moment'

150

4 Loanwords in OJ

4.3

Phonological adaptation

In most cases we do not know the exact shape of the proximate source of loanwords in OJ and it is therefore difficult to say much about the phonological adaptations that took place in the course of the borrowing. Some forms appear shortened or otherwise simplified in comparison with the distal source, e.g. OJ uru- from Skt. vnni, but it is not possible to know where in the course of transmission of this word the abbreviation took place. In particular, if a word passed through Chinese it would have been made to conform to its largely monosyllabic morpheme structure, cf. Skt. piitra- giving OC *pat. It is easy, however, to see that some adaptation to the simple CV syllable structure of OJ took place, most conspicuously in the insertion of epenthetic vowels to avoid syllable final consonants. The addition of -i or -u to loanwords is well known from NJ, and OJ kinu and /ami are examples of that, as is perhaps also pati (although the :MK shape pali makes us suspect that the -i may predate the arrival of the word in Japan). However, this seems to have become more common from EMJ onwards. A more widely used strategy in the older loanwords was the insertion of a post-consonantal echo vowel to echo the preceding vowel, e.g. OJ kasa (cf. :MK kas), kama (OC *khaam), pakase (EMC *pak dl;i:'),para (:MKpel) are straightforward examples of this, but others include pyera <pre-OJ *pyara (:MK pyet), tera < *tyara (:MK tyel). REFERENCES General: Miller 1967, K. Sato 1982, T. Sato 1982, Umegaki 1978, Unger 2001, 2003. Korean: Bentley 2000, Kanno 1978. Chinese: Miyake 1997. Sanskrit Suzuki 1978.

5

Eastern Old Japanese

The language of the overwhelming majority of OJ sources is that of the capital Nara, or the area around it. However, some sources reflect dialects from eastern Japan, usually collectively referred to as Eastern Old Japanese. Volume 14 of the Man 'yo shu consists of 230 azuma uta 'eastern songs/poems', out of which 89 have the place of origin identified in the commentary. Most of these were recorded by people from the capital travelling to the east in various official capacities. Volume 20 includes 93 sakimori uta 'borderguard songs', composed by soldiers from the east, mostly serving on guard duty in Kyushu, again mostly recorded by scribes from the capital area. Finally, the Hitachi fudoki includes 9 songs in EOJ. This material does not reflect a single dialect, but at least three main areas which may be labelled: (a) northern EOJ, comprising poems from (or dialects spoken in) the provinces of Kazusa, Mutsu, and Shimotsuke~ (b) central EOJ: Hitachi, Kozuke, Musashi, Sagami and Shimosa~ and (c) southern EOJ: Shinano, Suruga, and T6t6mi. These dialect areas differ appreciably between each other, with southern being least and northern most different from the central OJ dialect of the capital area. However, the majority of the poems and songs which are identified as EOJ have no information about their provenance. The EOJ material reflects features of EOJ to varying degrees, presumably due to the fact that most of it was written down by speakers of the central OJ dialect who were not equipped - or inclined - to record all linguistic features in the EOJ poems they were writing down. It is thought that the azuma uta of volume 14 of the Man 'yoshil among the EOJ material reflect least systematically the language of the eastern dialects. For example, there are more examples of phonological differences between central OJ and EOJ in the sakimori uta than in the azuma uta. The quality and especially small quantity of this material mean that what we can learn about EOJ dialects from it is limited. However, it is the only material which to any significant extent reflects dialectal material among the pre-modem Japanese text corpus, and at least one identifiably EOJ feature is reflected in standard cNJ (the vowel base imperative, see 16.4). In the following we outline some main features ofEOJ, but do not attempt a full description or a characterization of individual dialect areas. 151

152

5 Eastern Old Japanese

5.1

Phonology

In phonology, the EOJ material does not exhibit the ko-otsu syllable distinction in syllables with front vowels. That is to say, although the man 'yogana used to write individual words in the EOJ texts can be interpreted as having the sound values they had when writing OJ, there is no systematic distinction between Ce 1 (/Cye/) and Ce 2 (/Ce/), or between Ci 1 (/Ci/) and Ci2 (/Cwi/). However, Co 1 (/Cwo/) and Co 2 (/Col) appear to have been kept distinct. At present no phonological distinctions in EOJ which are not found in central OJ have been discovered. If they existed, the small quantity of the material would make them very difficult to identify. Where central OJ contracted juxtaposed *-i-a to /-ye/ in some grammatical morphemes, EOJ gave /a/, e.g. the morphological stative *i-ar- > central OJ -yer- :: EOJ -ar-, or with the adjectival copula conditional: *-ki-aba >central OJ -kyeba :: EOJ -kaba. 5.2

Morphology

In morphology there are several well-known features ofEOJ which differ from central OJ. First, EOJ is usually thought to have had a distinction between conclusive and adnominal among the QD verbs, see the basic paradigm in (1) of yuk- 'go', which also shows that there was no distinction between exclamatory and imperative:

a- stem

Central OJ yuka-

EOJ yuka-

Infinitive Conclusive Adnominal Exclamatory Imperative

yuki yuku yuku yuke yukye

yuki yuku yukwo yuke yuke

(1)

This is based on examples such as (2a) below. However, the picture is not as clear-cut as it is often presented. First, the number of examples of the distinct EOJ adnominal form is very small, with many more examples of verbs in adnominal function, and the other functions associated with the adnominal form, ending in -u as in central OJ. Second, it is noteworthy that a few instances of the EOJ forms ending in -wo occur in contexts where usually a conclusive form would be used. Third, even in some songs which also have the distinct adnominal form in -wo, a form ending in -u is used in adnominal function, e.g. ikiduku in (2b), which also has the distinct EOJ adnominal forms sumo and kinwo.

5.2 Morphology

(2) a.

ikapo

no

153

nero

ni

purwo

ywoki

lkao GEN top DAT fall.ADN snow 'snow falling on the top of Ikao' (MYS 14.3423) b.

oki

ni

sumo

wokamo

no

mokoro

offing

DAT

live.ADN

small.duck

GEN

like

yasakadori

ikiduku

imo

wo

deeply

sigh.ADN

beloved

ACC

okite

ki-nwo

kamo

leave.GER

come-PERF.ADN

EMPH

'I have come, leaving behind my beloved who sighs deeply like the little duck that lives in the offing' (MYS 14.3527) The distinct EOJ adnominal is an important fonn for reconstruction of pJ verb morphology. It may be thought to reflect a pre-OJ ending *-o, which had raised fully to -u in central OJ (by mid vowel raising, cf. 2.7.2.3), but only partially to -wo in EOJ. It is therefore usual to reconstruct pre-OJ conclusive/adnominal fonnants on the basis of the EOJ fonns, as shown in (3). However, it is equally possible to reconstruct, as in (4), a single finite verb fonnant, pre-OJ *-o, which in central OJ exhibits full mid vowel raising (*-o > -u), but in EOJ has full (*-o > -u) and partial (*-o > -wo) mid vowel raising in different environments and is morphologized differently in different positions. 1 (3) Conclusive Adnominal (4)

Central OJ -u -u

< <

pJ *-u *-o

> >

p(re-O)J

Central OJ/NJ

EOJ -u

-wo EOJ -u

-u

<

*-o

>

-wo The adnominal of the adjectival copula appears in some cases as -ke, e.g. naga-ke 'long' :: central OJ naga-ki. This is thought to reflect an older shape, which is also attested in a single song in the Kojiki, which raised (again by mid vowel raising) *-ke > (pre-OJ -kye >)central OJ -ki :: EOJ -ke (- -ki). 1

Support for this hypothesis may be found in the fact that most occurrences of conclusive verb forms are found not in absolute sentence final position (which is one environment where partial mid vowel raising is expected), but before extensions and final particles which probably combined with the preceding verb to form a single phonological word, landing the verb ending in nonfinal position which is an environment which favours full mid vowel raising (cf. 2.7.2.3); this would explain why it is the conclusive which has the fully raised reflex of *-o. See further Frellesvig (forthcoming) about this.

154

5 Eastern Old Japanese

EOJ has a negative auxiliary, -(a)nap-, which is not found in central OJ. It has main forms as shown in (5), which appears to be a hybrid, defective conjugation (note that the exclamatory only is attested as a stem), but which clearly is verbal as opposed to the composite central OJ negative auxiliary ( cf. 3.1.4.5.1). An example is given in (6a). Other EOJ negative forms not found in central OJ are shown in (6b-c), labelled gerund and infinitive, respectively. (6b) also contains one of the few words that are found inEOJ, but not in central OJ, namely sida 'time' (which may be a loan from Ainu, cf. 4.1). In NJ, negative formation is a major isogloss which separates western Japanese dialects (including Kyoto) from eastern dialects, including standard cNJ (cf. 16.3), but it is not likely that the eastern and standard NJ forms reflect the EOJ negative. (5)

Base a- stem Conclusive Adnominal Exclamatory

(6) a.

tusima no ne pa sita-gumo ara-napu Tsushima GEN top TOP under-cloud exist-NEG.CONCL 'There are no low clouds on the peak of Tsushima.' (MYS 14.3516)

-(a)nap-napa-napu -nape -nape-

b.

ywopi na pa ko-nani ake-nu night you TOP come-NEG.INF dawn-PERF.CONCL sida kuru time come.ADN 'you won't come at night, but you will come when it has dawned' (MYS 14.3461)

c.

ura-gare se-nana treetop-wither.INF do-NEG.GER 'When the (leaves on the) treetops do not wither ... '(MYS 14.3436)

The imperatives of vowel base verbs attach -ro, as opposed to central OJ yo. The EOJ imperative is the only feature of distinct EOJ grammar which is directly reflected in standard cNJ ( cf. 16.4). (7)

Central OJ mi-yo

EOJ mi-ro

REFERENCES Fukuda 1965; Mizushima 1984, 2003; Russell 2006, forthcoming; Vovin 2005, 2008.

Part II

Early Middle Japanese

6

Writing and sources

6.1

Writing

From the late Nara period a gradual process of simplification and abbreviation of the shapes of man 'yogana took place, slowly at first but gaining speed through the early part of the Heian period, and eventually resulting in the development of hiragana and katakana scripts. As the reduced shapes were dissociated from the kanji from which they derived, they unambiguously became phonograms and it became possible to write Japanese phonographically in a simple and straightforward manner. From the beginning of the tenth centwy until the end of the eleventh century we have a large body of text which was written almost entirely phonographically in hiragana, with only a few logographically used kanji. This style of writing is a continuation of the tradition of extensive phonographic writing which developed in the context of writing or recording poetry in the OJ period. Most of the EMJ hiragana texts were written by women who traditionally were excluded from the world of learning and Classical Chinese. The first line of the Tosa nikki from 935 where its author, Ki no Tsurayuki, somewhat tongue-in-cheek pretends to be a worn an trying her hand at writing diaries, is famous: wotoko mo su naru nikki to ipu mono wo wonna mo site mimu tote suru nari 'I as a woman will also try keeping this thing called diary, which men are said to keep.' It also shows, however, that it is a stereotypical oversimplification to view prose writing in hiragana, or in Japanese, during the Heian period as the exclusive domain of women. Also both men and women wrote poetry in hiragana. 6.1.1

Kanji-kana majiribun

From the middle of the Heian period the style of writing known as kanji-kana majiribun ('mixed kanji and kana writing') started being used more widely. Kanji-kana majiribun has a heavy preponderance of kanji, used logographically for most content and some grammatical words, and with kana (almost always katakana), used phonographically for some grammatical words and endings. The earliest extant text identified as being written in kanji-kana 157

158

6 Writing and sources

majiribun is the Todaiji fujumonko from the early ninth century, but it was not until after the middle of the EMJ period that kanji-kana majiribun became more widespread and was used in particular in the setsuwa literature (6.2.1). From the LMJ period it became widely used and largely replaced hiragana writing in prose. Thus the writing of extensive text passages in hiragana (or katakana) did not survive the early LMJ period and, regrettably perhaps, the elegant and economic tradition of simple hiragana writing, which is a fully sufficient means of representing Japanese, was lost. It also is worth noting that the establishment and spread of kanji-kana majiribun from the late LMJ period coincides with the establishment of the written norm of Classical Japanese. The way Japanese is written today is a direct descendant not of the hiragana writing of the EMJ period, but of the kanji-kana majiribun of EMJ and LMJ, although the proportions of logographic and phonographic writing differ between then and now, with more words spelled out phonographically today and with hiragana replacing katakana. Alllogographic writing of Japanese derives from reversing kanbun-kundoku (the rendition in Japanese of a Chinese text) from reading to writing (see 9.1.2), but the link is particularly close for kanji-kana majiribun which arose as a transfer of the techniques of kunten (annotations to a Chinese text) to the writing of Japanese. Especially the practice of writing grammatical elements of a Japanese rendition in kana glosses next to Chinese text served as the model for the kanji-kana majiribun style of writing. In 9.1.1 there is a short example of an annotated Chinese text and it is conspicuous that its appearance is quite similar to Japanese texts written in kanji-kana majiribun. Also the Japanese language used in kanji-kana majiribun was heavily influenced in style, vocabulary and grammar by the Japanese language used in kanbunkundoku, the so-called kuntengo (9.1.6). 6.1.2

Kana (hiragana, katakana, hentaigana)

The hiragana and katakana developed as reduced shapes of man 'yogana. Table 6.1 shows the standard sets of kana and their kanji origins in the usual 'fifty-sound table' (gojuonzu Ji. +if~, see further 6.1.4) grid arrangement, in ten columns by five rows plus an extra slot for the moraic nasal. The columns (gyo 11') are referred to by the head kana, e.g. a-gyo 71f the 'a-column' or ka-gyo jJ:ff the 'ka-column', and the rows (dan.®: (or retsu ~IJ) by the single vowel kana, e.g., a-dan 7 .®: the 'a-row' or i-dan -{ .®: the 'i-row'. Of these, the following kana from the wa-column were abolished in the writing of modern Japanese in the script reform of 1946 and are today only used in the writing and presentation of Classical Japanese: l;;J/ 4, 2.1 .7.., :a:I 7 (except that the latter pair is used for the accusative particle o). As may be seen, in some cases equivalent hiragana and katakana originate in the same

Table 6.1 Kanji origins (!(kana

a

ha iP 1JD ~ {ii: f;:_ /t t:t. ~ ti rEt tJ 1JD -if ~ 7 ~ 7 ~ /"\ J\ ki si ti ni hi ka

~

~

7

~iiJ I

sa

v:~

PA

~

~

L

1

fjt

~

~

y

u

ku

>

I

~

<

7

~

7

e

z ~ z 1su

-t

rJi se

;t

ti.

~t

.:r.

ri

m- -tt- tlt

'T

~

to 1t

.,. '-

?.t 1t

:J

tlt

-l!

ko

0

so

a a

~ ')

~

=f tu

J,. '"9 J,. .A

ke

na

ta

if if

t;:

{-

--

0

{-

l::

.It .It

rna

** mi*

7

lj. .... ..... '-

ra

ya ~

~

~

-y

~

7

ri

~

t)

--

l)

~

wa ~

b

~

?

ltl ltl wi

flj ~ ~ fU 4 #

nu hu mu ru yu "'? Jll ~ "9$. ~ ~ tJ jt $ E8 0 il :J_ 'Y 1'N 7- "9$. 7 ~ b. $ E8 Jv il he re te ne me we A, $ liJ !7:. --c tJ. n *L JJ,. -m A, $ ;) !7:. v tL .7. l{ 7 no ho mo ro wo to N yo ::g g c ll: (]) l!J tl 1¥ t ~ ~ 0 ~ ~ lv 7G r ll: J l!J 1¥ .::c ::g 3 ~ p g 7 .p. /

x *$ x ;;.. *$

*

160

6 Writing and sources

kanji (e.g. ka, na, ma, ya, ra, wa), in others not (e.g. a, sa, ta, ha). Overall, the reduction took place in two ways: as a general rule, hiragana resulted from a cursive writing of full man 'yogana, whereas katakana originate in parts of man 'yogana. It must be emphasized, however, that man 'yogana in their full shape continued to be used through the MJ period as phonograms. It is not the case that hiragana or katakana simply replaced man 'yogana: they coexisted for a long time. Published editions of texts today use the standard sets of kana, but that inventory of graphs was only settled upon in the Elementary School Order (ShOgakkorei) of 1900. Through the MJ period a large number of competing and variant, but phonographically equivalent, shapes of kana was used~ these are today known as hentaigana. Table 6.2 below gives an impression of the variability among early kana shapes. Legend ascribes the invention of hiragana to the priest KUkai (774-835), but it is clear that kana were not invented by any individual or group. Many man 'yogana are graphically complex and cumbersome to write, and it is only natural that they would tend to be simplified and abbreviated for ease of writing. With the discovery of more primary material from the OJ period, mostly in the form of mokkan (wooden tablets, see 1.2.2), examples of graphically reduced man 'yogana are attested in increasing num hers also from the OJ period. It is important to note that many of the reduced shapes which became hiragana in Japan are found in Chinese cursive writing styles, and that many of the reduced shapes which became katakana are found in Korean kugyol (cf. 9.1.1). This suggests that in fact the hi ragana and katakana letter shapes were not the result of independent developments in Japan, but followed continental models. It is only possible to follow the gradual emergence of the hiragana and katakana letter shapes through authentic, contemporary writing, not later copies of manuscripts. In addition to mokkan and fragments of various sorts, many of which seem to be incidental writing, or even writing practice, the most important authentic, primary text materials from the EMJ period are annotations on texts, the so-called 'kunten' materials (see 6.2.2). The study of the development of kana, especially katakana, and of dakuten (below) is intimately linked to the study of kunten materials. There exists a large amount of such materials in the possession of Buddhist temples in Japan, exhibiting great diversity in traditions of annotation, including letter shapes. Tsukishima (1986) reproduces in several hundred tables examples of kana shapes and other writing conventions drawn from such kunten materials. In Table 6.2 (taken from Tsukishima 1977: 98) are shown kana shapes (and a few other annotation conventions) from annotations (dating from 883) of the Dizang shilim jfng (:t-1!!~+~*£, Japanese Jizo-jurin-gyo), a Chinese translation of the 'Sutra of the Ten Cakras of K~itigarbha'. Even a cursory

6.1 Writing

161

Table 6.2 Early kana shapes (fi-om Tsukishima /977: 98)

1 7

--

/

?

LJ

"""') lv-. v,

..../

7

'* ' \,."l

1

~

..,. ?

1-tft 4

7

r

7- 3' -lj- 7J 7 ,~ -1:! 7) ' ~ ~\ ~ 1l2 ~' tJ :; n J) j 1' *7 fi -1: ~ ~~ ; "')

~ -::? ~' I \

/;)

?:..

l)

~~ ~ 'J

1-

=>

·~

::1

~

r}

""' -t-I) ~

=>" 1-

Jv •L ~

:;...

N?

.::L 'ci) \~

~ \:l

~ ~

~ l -;}"

T -t

v

-

-

.t -

T ?) ~ 0

A

7

" ,,

L,.. b

L-

I}

1 1:

-. A 7

~

7 0 3 -1:: -t ~ A -5 -c. .. A

.#

-t

..'...""'...

~ ~' a; l7 r~

~

....

~~

~

~

~r.

;...

J

:J. ''/ 'A ~ ry

~ ;l ~ ~

J ........... t ~L ~ Jl. '2 + ~ )( \. •..::..' L I .) ... , ";J. :L

~

. ? y t: ..... 1- ... ~ 1 ;k. - t, .l t· 1;> ~ L 1' p .... .....

*

-*

-"?

'

~

7

'? 1-...

~

~

cJ "(i:.. l "r,,. ~ ~ h T"t ~

J

~~... 1)

t 1~ ....

I

~

~~ h <

~~

fl)

J

7

L-

"

'/

J: ~ ~

}-1)

7

~

;t<

~ .. ~

~

~

......

::1

~

t,.. ~ ... h L ....

._

t-

~

inspection gives a clear impression of the diversity and also reveals that both hiragana- and katakana-like letter shapes are found in a single set of annotation. It is important to emphasize that all the letter shapes in Table 6.2 are from a single set of annotations and do not represent different steps in the development of letter shapes.

162

6 Writing and sources

As such materials are further studied and published, the details of our knowledge about the development of the use of hiragana and katakana will change, but the general picture is as explained in this section. The different strategies of reduction eventually resulted in two distinct sets of kana, hiragana and katakana, which were associated with functional differences: Parts of man 'yogana, which yielded katakana, were practical to use in annotations on texts where space is limited~ in addition to that use, katakana also later became used in kanji-kana majiribun. Throughout the premodern period katakana maintained some connection with kanbun or writing incorporating kanji. Cursively written whole man 'yogana, eventually giving the modern hiragana, were thought aesthetically pleasing and were associated with literature and calligraphy. The hiragana were also practical for writing running text, as strings of letters could be written without lifting the brush from the paper. When talking about writing in the Heian period, hiragana has become known, and was also at the time occasionally referred to, as 'onna-de' ('woman's hand'). Whereas academic, intellectual or public writing was mostly done in Classical Chinese (or in hentai kanbun, see 9.1.2.1 ), personal and private writing in Japanese in hi ragana was associated with the leisure of the (female) aristocracy and as is well known, much of the prose literature in Japanese from this period was written in hiragana by women. One exception was composition and writing of poetry in hiragana, which to some extent was public and in which men certainly took part. As noted above it is wrong to view writing in hiragana, or in Japanese, during the Heian period as the exclusive domain of women. It must further be noted that just as there was a gradual transition and long period of coexistence between kana and man 'yogana, so the letters which we today identify as hiragana and katakana were not functionally differentiated from the outset, or conceived of as belonging to different sets of letters~ thus among the hentaigana we find katakana(-like) letter shapes, and conversely the kunten materials also hold many hiragana(-like) letter shapes mixed in with katakana, see Table 6.2. The hiragana and katakana fore illustrate this well: they originate in different man 'yogana which represented distinct syllables in OJ and early EMJ, ;t < t< (OJ I.e!) and ::c. < ii (OJ /.ye/), and they can both be found within single bodies of text. Overall, however, the later functional differentiation between hiragana and katakana has its origin and basis in the two different strategies of simplification and tendencies in use. 6.1.2.1 Sei'on and daku'on In OJ, different man 'yogana were generally used for sei 'on and daku 'on (i.e. syllables with initial tenuis (/p, t, k, sf) or media (!b, d, g, zf), respectively~ see 2.2.2). Thus, for example, the syllables ka and ga could be written with different characters (e.g. ka JJO~ ga ~). However, even in OJ this phonological distinction was sometimes ignored, and for example a man 'yogana normally

6.1 Writing

163

used for ka could be used for ga (e.g. J.JO sometimes used for the genitive particle ga). In early EMJ the sei-daku distinction came to be consistently ignored in general writing and the same man 'yogana were used for pa/ba, tal da, ka/ga, salza, etc. Accordingly, as the abbreviated hiragana and katakana forms developed, separate letters for sei 'on and daku 'on did not evolve and this phonological distinction remained unnoted through most of the MJ period in general writing. Thus, although the tenues and mediae clearly were phonologically distinct in MJ, as in OJ and in NJ, the kana letters from the pa, ta, ka, and sa-columns were used to represent /pV, bVI, /tV, dVI, /kV, gVI, and /sV, zVI, respectively, such that for example t'i was used for both pa and ba, t-c. for ta and da, iJ., for ka and ga, and ~ for sa and za. This is reflected even today in the absence of separate kana letters for sei 'on and daku 'on, with daku 'on being indicated by adding a diacritic, the dakuten, ·,to the top right comer of kana for sei 'on, e.g. iJ., ka, ti). ga. Such a development towards a more underspecifying script is not unique, but paralleled, for example, in the transition from the older (Germanic) to the younger (Scandinavian) runes (completed by the middle of the ninth century). As opposed to the older runes (with twenty-four letters), the younger system of sixteen letters did not have separate letters for tense (unvoiced) and lax (voiced) stops, in addition to other instances of equivalence. Thus there was in the Scandinavian runes only one letter for /p, b/, It, dl, and /k, g/, respectively, although these sounds were and remained distinctive in the Scandinavian languages and had had separate orthographic representation in the Germanic runes. It is not difficult to understand that native readers have few problems with a phonologically underspecifying script in general writing, as they know the words in the language and can guess from context which words are meant, even in the absence of an absolute phonological identification, in the same way that phonologically or phonetically underspecified speech is usually readily understandable. Thus, for example, most scripts have no expression of suprasegmental features such as accent or tone. Note finally that although tenues and mediae have been phonologically distinct through the history of Japanese, a sizeable proportion of mediae derives from tenues either in morphophonemic alternations as a result of rendaku (see 2.6.2), e.g. sakura 'cherry' + pana 'flower' => sakurabana 'cherry flower', or by automatic phonological neutralization of tenues as mediae after nasals (cf. 7.1.2.2), e.g. (yom- 'read'=>) yoN-+ -te => yonde 'read.GER'. In such cases no ambiguity in morpheme identification arises, even if the phonemic shape is not unambiguously noted. 6.1.2.2 Dakuten In specialized writing, however, various means were used, when it was thought necessary, to give a more precise indication of the phonological shape of a word, noting whether a syllable was sei or daku, or its phonological pitch.

164

6 Writing and sources

This was mostly done by diacritics, but there are also cases of, for example, inverted kana being used to specify a daku 'on. The earliest attested use of diacritics to mark sei or daku on man 'yogana is from the late ninth century, and on kana from the eleventh century. Diacritics include for example 1 , short for ~(if) daku(on) (or the verb nigoru ~'pronounce as a daku'on'), for example ~ ~ to specify ba (and not pa). Usually, however, diacritics were dots, circles, lines, even triangles, or combinations of these. They could be added to man 'yogana or kana, but they were also added to logographically used kanji to give a hint to their reading. EMJ texts in which sei-daku (or pitch) was noted were almost exclusively annotations (kunten materials) or dictionaries, and sei-daku thus remained un-noted in general writing in the EMJ period. During the early LMJ period (Kamakura), annotation of EMJ texts in Japanese to specify features of pronunciation along those lines became current, but it was not until the NJ period, from the beginning of the Edo period, that the dakuten we know today became established and widespread in general writing. The dakuten eventually settled upon appears to have had two sources: (Chinese) tone dots and the letter anusvara from the Siddham script. In some documents, including the dictionary Ruiju-myogi-shO from 1081 (6.2.3), seidaku was noted in conjunction with pitch by means of tone dots which mark tone or pitch by their position next to a kanji; see 7.4.1. When extended to note sei-daku, the tone dots indicate pitch by position, while their shape indicate sei-daku, usually using single dots or circles for sei 'on, but double dots or circles for daku 'on. However, separate from the notation of pitch was a diacritic of the shape ~ which in early annotations was placed next to or below a letter, but later customarily on the right hand top corner, the position in which the present-day dakuten is placed. This diacritic is thought to have its origin in the letter anusvara from the Indic script Siddham. The Siddham script was used to write Sanskrit and was brought to Japan in the early ninth century, most likely by the priest KUkai when he returned in 806 from his study tour to China where he also studied Sanskrit. Siddham was, in India and other places, later generally replaced by the Devanagari script to write Sanskrit, but Siddham has remained in use in Shingon Buddhism in Japan, mainly for copying out sutras and mantras. The letter anusvara represents a nasality feature of Sanskrit. It is usually romanized as -rrz, but in Siddham it appears as a dot above a syllable, e.g. sa -'< versus sarrz -ii. The diacritic • was used with kana, man 'yogana, or with kanji used logographically for Chinese words. In addition to marking daku 'on, it is also in some annotations used to mark syllables with an initial nasal (n-, m-); there are also cases of it being added to kana for .u or .ito note a nasalized vowel [fi] or [i] which were renditions of Chinese 1-TJI, e.g., ~ [fi] or ~' [i]. This diacritic was in other words used mainly as a nasality mark;

6.1 Writing

165

this provides further evidence that the mediae phonetically were pre-nasalized in EMJ (see 2.2, 7.1.4.3, and 11.1) and were associated with the phonetic feature of nasality also in metalinguistic consciousness. 6.1.2.3 Handakuten OJ /p/ was lost in most contexts in the course of sound changes which occurred through the MJ period, merging with /-w-/ in medial position after vowels (but preserved after /Q/) in EMJ (7.3.1.2), and changing to If-! in initial position in LMJ; see 11.3 for details and exceptions. The limited contexts in which /p/ was preserved were readily recognizable and a specific way of writing /p/ as distinct from /f/ did not develop in the MJ period. However, towards the end of the LMJ period Portuguese Jesuit missionaries instituted the use of a circle on the top right comer of a kana for contemporary jV (present-day hV) in order to write unambiguously pV, e.g. t'i ifa/ha), 1'1 (pa). This was first used in the Rakuyoshii, a kanji dictionary published by the Jesuit press in Amakusa in 1598. Since then this practice gradually spread and is, of course, today a fully integrated feature of Japanese writing. 6.1.2.4 Orthographic categories; the Iroha-uta, the Japanese 'alphabet' Although there was a large inventory, including variant shapes, of man 'yogana and early kana letters with widespread phonographic equivalence (i.e., many different letters for the same sound), the establishment and awareness of distinct orthographic categories is evident from three mnemonic word lists or poems from the first half of the Heian period in which each distinct letter category occurs only once: the Ame-tsuchi no kotoba (or Ame-tsuchi no uta), Taini-uta, and 1roha-uta (see (1)-(3) on pp. 166-7). These lists/poems functioned as a kind of ABC to remember the distinct letter categories and were also used for writing practice. The Ame-tsuchi no kotoba is basically a list of words, but the Taini-uta and 1roha-uta are organized into phrases or clauses conforming to some extent to the native poetic metre of alternating seven- and five-mora-long verse lines. The Ame-tsuchi no kotoba has forty-eight letters, but the latter two have the forty-seven kana letter categories which, with the later addition of /vi'/, are in use today. The three lists concur in not having distinct letters or representation for sei 'on and daku 'on. Chronologically, they all reflect a stage of the language after them erger of the kO-otsu distinctions (as there is no representation of these distinctions), but before the merger of non-initial /-p-/ with /-w-/ (950-1000, see 7.3.1.1), as some pV kana are illustrated in positions in which /p/ later merged with /w/, e.g. ii•I'J: kapa 'river' (> kawa) distinct from 19:>:b yuwa 'sulphur', as well as before the merger of /.wo/ and /.o/ (c. 1000, see 7.3.2.3), as both :a:- wo and :t3 o find distinct representation. The Ame-tsuchi no kotoba alone reflects a stage of the language prior to the merger

166

6 Writing and sources

(1)

Ame-tsuchi no kotoba

0 ~~ L, .:c 1::, ~ "i tp. l'i ]f. tJ, t-c. ';: < IJ u ;s :. It V' (: v\ lb. 5 ""' T % ~ b .to b it J: *- (]) *- ~ t~ tL ~ -c ~0 ~~ G .:C 1::> ~"i iPI'i ]f.td,

&>; II) ~

t

~

~

~

&>; ~

arne

tuti

posi

sora

yama

kapa

mine

heaven

earth

star

sky

mountain

nver

peak

td;: tani

~I?

U ;s

:. It

'(} (:

kuma kiri

muro

koke

pita

v\lb. inu

valley

cloud mist

room

moss

man

dog

upe

suwe yuwa

saru

opu

seyo

above

end

sulphur monkey grow

do!

e

no

ye

wo

nare

wite

branch

ACC

get used to sitting

itJ:

?""'

hackberry tree GEN (2)

Taini-uta

t-c.

';:

-c

u

-t (:

~ v\ t~ ~ b tL ~ .:C ~ h- lh (b~IJ.to'U'~<~"iL,;s(])?0%""~ :. ~::> t 1-x ~~ it J: b tJ. n~ It lb.

*-

t-c.~

1;:

v\-c

t~

~u

btL

~

.:c

tawi

ni

ide

na

tumu

ware

wo

zo

paddy

DAT

go out

greens

pick

I

ACC

FOC

~ h-

~T

(:

.to'U'~<

kimi

mesu

to

asari

opi-yuku

you

see.RESP

when

hunt

chase-after

? 0%""~

:. 1::,

yamasiro

no

Yamashiro

GEN

utiweperu kora is very drunk dear.girl

~"iG;s

tl'i

~~it J:

*-

btJ.

mopa

poseyo

e-

pune

n~lt lb. kakenu

seaweed

dry!

(un)able-

boat

doesn't anchor

6.1 Writing (3)

167

Iroha-uta

v\ ;s l'i

~;:

..:c """'.)

h ft. t:;

-c it;

~

v\;s

I'X

~

ira pa colour TOP

:b

fp.

.r-...

~~

~

u ?

19:>/i;JJ.r.

i::J

~

lb.

~(J);B

o :a:- :b < ~~

l..-%0 'bitT

tc tL It h :. *-

t.P J:

l:.f~.r-...~

i::J~Ib.o

:a:-

nipopedo although shines beautifully

tirinuru will scatter

wo

ten ..:c

J:

wa ga yo I GEN world ?~

uwi material. world it;~~

19:>11)

asaki shallow

yume dream

tare who

"""'.);h tune always

zo FOC

EXCL

ft. t:; u naramu will be

(J)

:B<~~

lth

:.;;t-c

no

okuyama deep.mountain

kepu today

koete crossmg

GEN

J.r-1__., mizi won't see

2.0

t

wepi mo get drunk ETOP

itT sezu doesn't

'Colors are fragrant, but they fade away. In this world of ours none lasts forever. Today cross the high mountain of life's illusion [i.e. rise above this physical world] and there will be no more shallow dreaming, no more drunkenness [i.e. there will be no more uneasiness, no more temptations].' (Translated by Andrew N. Nelson (1974: 1014)) of the syllables /.ye/ and I.e/ (before 950, see 7.3.2.2). The Ame-tsuchi no kotoba is not attested until the Minamoto Shitago-shil, compiled 967, in hi rag ana, but it has two occurrences of hiragana ;;t e, interpreted as representing I.e/ and /.ye/, respectively, as there is no other good explanation for the double occurrence of X.. The current kana inventory of course does not include representation of that earlier distinction, but in addition to man 'yogana, kana annotations from the ninth and early tenth century have distinct letters for I.e/ and /.ye/, shown under t<. in the a-column and under ti in the ya-column in Table 6.2 above, and as mentioned the current hiragana and katakana in fact derive from early kana for I.e/ and /.ye/, respectively: ;;t < t<. I.e!, while .:r. < ti lye!. The Taini-uta is attested first in 970 (in the Kuchizusami, a first primer for the education of boys of the nobility, compiled by Minamoto no Tamenori, ?-1011), which is probably fairly close to the time it was made, but the first attestation of the Iroha-uta is in the Konkomyo saish88 kyo ongi from 1079, although it must have been made earlier. Nothing is known about the

168

6 Writing and sources

authorship of the Iroha-uta. It has, like the hiragana script itself, been attributed to KUkai, but the phonological system reflected in the Iroha postdates Kiikai, who died in 835, by a century and a half. 1 The Iroha-uta soon superseded the Ame-tsuchi no kotoba and Taini-uta, which today are not generally known, and from the late Heian period became widely and commonly used to remember the kana categories and for writing practice. It is difficult to exaggerate the importance of the establishment and dissemination of the Iroha-uta. It is a defining event in the history of Japanese script and writing, and more generally in Japanese culture, linguistic and otherwise, and for example played an essential role in the spread of literacy in Japan through the LMJ and NJ periods. The Iroha-uta finally defined the forty-seven kana letter categories which, with the sole addition of lv// for the moraic nasal, still are in use today, enshrined the principle of not having separate letters for sei 'on and daku 'on, and fixed the sequence of the letters in a list. Phonographically, the Iroha inventory may be seen to have some shortcomings, viz. the systematic underspecification of the sei 'on-daku distinction, and a lack of distinct, single-letter representation of 'new' sounds in the language (see 6.1.2.5), so that as the Iroha letter categories became set in stone, the relation between sound and written representation over time was obscured by sound changes (see 6.1.3). However, the Iroha represents a good phonological analysis of the number of (free) moras in the language at the time it came into being, and it successfully established and fixed a small manageable inventory of letter categories, thus providing a writing system which was learnable for larger groups of people. The fact that the letter categories of the Iroha by and large are those used to write Japanese today is testimony to its success. In addition to establishing the letter categories and being used for writing practice, the Iroha-uta also in more general cultural terms assumed the functions of an alphabet or ABC. It came to be used as a sequence for enumeration, and is still today sometimes used in lists, or in subclassification within dictionary entries. One enormous impact of the Iroha was in its use for ordering ('alphabetizing') dictionaries from as early as the end of the Heian period (used first in the Irohajiruisho compiled between 1144 and 1181), making possible the compilation of dictionaries with easy, phonologically based lookup. The Iroha continued to function as the Japanese alphabet, as a linguistic and meta-linguistic organizing principle until it eventually was replaced by thegojuonzu in the middle of the Meiji period (6.1.4). The wordlroha (written variously as v\61-J:, ffr§~. -§..)also came to be used as a name for the 1

It is interesting to note that iropa is a word meaning 'natural mother' which could be taken in the sense of 'origin, beginning', here of writing. The similarity between the words alphabet and

iropa is surely fortuitous.

6.1 Writing

169

kana inventory and by extension to mean 'ABC' in the sense of 'rudiments, basics (of something)', cf. living expressions such as iroha no 'i ' no ji mo shiranai (v' iS l"i (J) I{Jt J (J) t .*P ~ft. v') 'be utterly illiterate, or ignorant; lit.: not even know the {Jt [first] letter of the Iroha'.

*

6.1.2.5 New sounds In the transition between OJ and EMJ and during the EMJ and LMJ periods, a number of sound changes occurred. Some of these were mergers, or loss of phonological distinctions, for example the loss of the ko-otsu distinction, resulting in fewer orthographic distinctions being necessary. Others, however, introduced entirely new phonemes into the language, all related to syllable structure: bound moras (6.1.2.5.1), complex syllable onsets (6.1.2.5.2), and Sf syllable final -t (6.1.2.5.3). As the kana letters were reduced forms of man 'yogana, which reflected the phonology of OJ, they did not readily provide for the representation of sounds which were new to EMJ, so although some of these new sounds were present in the language at the time the Iroha was established, the letter categories of the I roha do not include distinct representation of these new sounds.

6.1.2.5.1 Bound moras Through the onbin sound changes (7.1.4), Japanese became quantity sensitive and acquired long syllables which included bound moras (7.1.1): moraic consonants IC, Q, Nl and moraic vowels', both oral II, U/, and nasal If, U/. Of these only the moraic nasal IN/ eventually got its own letter, the only addition to the letter categories of the Iroha. Various means were employed to write the moraic consonants IC, Q, Nl which appeared in the language in the transition between OJ and EMJ: Apart from not being represented (e.g. moQ-te ( < motite) 'holding; hold-GER' written as .::e- T mo. te (in annotations from the late ninth century) or siN-zi (< sini-si) 'dead; dieSPST.ADN', written (in the Tosa nikki, 935) as G G si.si), they have been written with a number of different kana for CV syllables. There was some overlap - within individual texts - in representation, and for example man 'yogana -$- (mu) or its reduced katakana shape A have both been used to represent both Nand Q; this reflects the phonemic identity between Nand Q (as /C/ unspecified for nasality) in morpheme-internal position ( cf. 7.1.2.1). The current use of kana for tu (~) to write the moraic obstruent /Q/ was originally inspired by the kana transcription of syllable final -t as ~ (tu) (cf. 6.1.2.5.3) and was settled upon relatively late. The convention of writing this letter in small size dates from the script reform of 1946. The hiragana letter lv for the nasal moraic obstruent IN/ is thought to derive from the man 'yogana 7C (mu), though some scholars believe it to be a further development from (a precursor of) katakana A (itself<-$-). Table 6.2 above shows two letter shapes for IN/, one like present-day hi ragana lv and the other

170

6 Writing and sources

seemingly a precursor of katakana /.However, the katakana letter/ seems to be first attested as such in the eleventh century. It appears to have no recognizable source in a man 'yogana, but to have been invented specifically for the purpose of representing IN/. The dot in / may well reflect the nasality diacritic • which was also a source of the dakuten (6.1.2.2). The bound moraic vowels, /1, U, i, U/, emerged through the onbin sound changes, and also came to be used in SJ loanwords. These sounds were written with kana for the free moras li, u/, i.e. v\ ? , with no differentiation of oral and nasal, apart from the use in some annotations of the nasality diacritic • with kana for .i, .u to specify If, U/ in renditions of Chinese /-rjl.

6.1.2.5.2 Syllables with complex onsets Syllables with complex onsets, consisting of consonant+ (palatal or labial) glide, represented in the moras Cya, Cyo, Cyu and kwa, gwa, kwe, gwe (7.2), entered the language in the EMJ period through the large-scale adoption and adaptation of SJ vocabulary. Somewhat later Cya, Cyo, Cyu also arose outside SJ vocabulary through sound changes. Also these sounds were not written with distinct single letters, but instead as combinations of kana for Ci or Cu with yV and wa, i.e., Ciya, Ciyo, Ciyu (e.g. {!s ~ kya, {!s J:: kyo, {!s 19:> kyu) and kuwa, guwa, kuwe, guwe ( :b kwa, <':b gwa, < % kwe, <'% gwe), respectively. Since the script reform in 1946 the glide initial letter has been written in smaller size: {!s ~, {!s J:, {!s 19>; b, <'b.

<

<

6.1.2.5.3 Syllable final /-t/ A sound found only in SJ loanwords was syllable final /-t/ (cf. 7.1, 7.2). This remained pronounced as 1-tl, i.e. without a following vowel, until well into the NJ period (see further 11.4). In annotations it was noted in a number of different ways, but in general writing it was transcribed by kana ----::> or sometimes 'G. It has recently been shown, however, that some LMJ sources made an orthographic distinction between 1-tu/ and 1-t/ by means of variant kana (hentaigana) which were originally used as equivalents for /tu/. In NJ final /-t/ acquired an epenthetic vowel to give tu or ti. 6.1.2.5.4 Recent Modern Japanese loanwords Finally, we should note 'new sounds', and especially new combinations of sounds, which have come into the Japanese language in the contemporary period, since the Meiji restoration, through the intake of large numbers of western loanwords (see 14.6). Like moras with complex onsets, the moras containing these sounds have been written with combinations of kana, e.g., /Jln 7 7 /, paati.i /-\-7-1'party', sherii "/' :r. ~ - 'sherry', tsaitogaisuto '/ 7 -1 !- jj -1 .A !- 'zeitgeist' ,ftrii :/ :r. ~- 'jelly', vandaru-jin ij 7 / !l Jv A 'Vandals (Germanic

6.1 Writing

171

tribe)'. The moras /wi, we, wo/ have reappeared in the language, but they are today written as wirnsu ? -1 Jv .A 'virus', wesutan ? ::c. .A ?I/ '(a) western', wokka ? ?.!"" v j:] 'vodka', rather than using the old kana for these moras (*4 ;v.A, *:x..A?t ::.---, *7 v j:}).

6.1.3

Orthographic norms

After the establishment of the orthographic categories in the lroha, a number of sound changes took place through the MJ period which rendered several originally phonographically distinct kana categories equivalent, in some cases unconditionally and in others only in word non-initial position: /-p-/ merged in intervocalic position with 1-w-1 as 1-w-1, but remained /-p-/ after /Q/ (7.3.1), and /w/ was lost in both word initial and non-initial position, first before /o/, and later before li, e/ (7.3.2.3). These changes are often stated as syllable mergers, as shown in (1) which lists the merged syllables, the outcome of the merger, and the previously phonographically distinct kana categories which could now be used to represent the outcome of the merger. Sound changes resulting in kana equivalence

(4) a. b. c.

In initial position /.of, /.wo/ /.if, /.wi/ /.e/, /.we/

> > >

/.of /.if /.e/

d. e. f. g. h.

In non-initial position 1-.o/, 1-.wo/, !-pol 1-.i/, 1-.wil, /-pi/ /-.e/, /-.we/, /-pe/ 1-.wa/, !-pal 1-.u/, 1-pu/

> > > > >

1-.o/ 1-.i/ /-.e/ 1-.wa/ 1-.u/

:B:a:-

v\ ih

;;t%

:t3

:a:-

f~

v\ ih vX.%-'"' :b f'J:

?b

These sound changes had two consequences for the representational value of kana: First, the kana categories :B:a:-, v\fh, and X.%, respectively, became phonographically equivalent in all contexts. Second, the representational value of the p-column kana changed in non-initial position, but the fact that there was no orthographic distinction between sei 'on and daku 'on and that the kana from the p-, t-, k-, s- columns were used to represent both sei 'on and daku 'on (lpV, bVI, /tV, dVI, lkV, gVI, and /sV, zVI, respectively) complicated the situation further. In initial position, as before, f'J: could represent /pa/ or /ba/ (mainly in SJ loanwords which had no restrictions on the occurrence of word initial media, see 7.2), and likewise "(} could stand for /pi/ or fbi/. However,

172

6 Writing and sources

in non-initial position l'i could now represent /wa/ or /ba/, or, after /Q/, /pal. Likewise, t} came to represent li, hi/ (or /pi/), .ro... could represent /e, be/ (or /pel), li /o, bo/ (or /pol), and b /u, bu/ (or /pu!). The result of this was partial phonographic overlap, but not full equivalence, of non-initial li, 0, .ro..., l'i, and b with :BI::a:-, v Wh, *-.1 %, :b, and ? , respectively. See (2) which shows the phonographic value of l'it}b.r-...li, and the partial phonographic equivalence with other kana in non-initial position: (5)

Initial position

l'i t} b .ro...

li

/pal

lbal

/pi/ /pu/

lbul

/pel /pol

lbol

/bi/ /be/

Non-initial position /ba/ /wa/ (= :b) /bi/ lil(=v\ ~) /bu/ lui(= ?) /be/ lei(=*-.,%) lbol lol (= :B, :a:-)

/Q//pal /pi/ /pu/

/pel /pol

Thus, for example, kawa 'river' could now be written iP:b or iPI'i, whereas kaba 'Japanese white birch' would only be written iPI'i. Likewise, kai 'shellfish' could be written ir>v \ iJ•~, or iJ•t}, whereas kabi 'mildew' would only be written as i:PV.. Conversely, from the reader's point of view, iP:b spelled kawa, whereas ir•l'i spelled kawa and kaba; and iPv \ and f.p~ unambiguously spelled kai, whereas iJ>t} spelled kai and kabi. Taking account of the spelling of words with 1-Qp-1 complicates the situation further. For example, kappa(-to) 'with a thud' could also be spelled iJ>I'i (depending on how /Q/ was represented). In (6), the spellings of some words which illustrate the polyvalence are listed, showing the potential difficulties for both readers and writers: (6)

kawa 'river' kaba 'Japanese white birch' kappa(-to) 'with a thud'

iJ>I'i iJ>I'i iJ>I'i

kabi 'mildew' kai 'shellfish'

iJ>t} iJ>t}

kae 'change (tr.); inf.' kabe 'wall'

iJ>.ro... iJ>.ro...

io '500' ibo(ziri) 'praying mantis'

\j\fi \j\fi

kau 'change (tr.); concl.' kabu 'head'

iJ>b iJ>b

6.1 Writing

173

Clearly, for the kana categories :to :a: l'l., v'C9 0, ;t 2. ""-, ? b, :bi'J:, the mapping between sound and writing became much less straightforward than it had been at the time the kana letter categories were established and they more closely reflected the phonology of the language, and as may be expected, this led to efforts to regularize the use of these kana letters. Rather than reforming the inventory and use of kana letters along simple phonographic lines (for example, by dropping the letter categories :a: C9 2. altogether and specializing non-initial I'J:Ob-'"'l'l. to represent lb VI and /p VI after /Q/), an etymological spelling principle gained ground from around the beginning of the thirteenth century, eventually resulting in the so-called rekishi-teki kana-zukai ~5!:.1¥J {litl. ifv ') 'historical kana-usage, historical spelling', which is still in use today in the presentation of premodern Japanese texts and in writing Classical Japanese. The simple principle is that any word is to be spelled the way it was, or would have been, spelled within the kana categories of the Iroha, before the sound changes outlined in ( 4) took place. Thus kai. 'shellfish' ( < OJ kapi) is spelled fp:Q_, ai 'indigo'(< awi) is
174

6 Writing and sources

li-t!>tP·, cf. mi 'body' andlika- 'close' or kana-zukai 'kana-usage' fp.ft--5tJ.,.

v\

cf. tsukai 'usage'). The etymological spelling principle was first explicitly proposed by the poet and scholar Fujiwara no Teika (1162-1241) as part of his recommendations of the spelling of individual words set forth in works such as Gekansha (?1241, a book on poetics) or Hekiansho (1226, annotations on poems from earlier poetry anthologies). His primary concern was the production of faithful editions of earlier texts, preserving their original shape, and he based his spelling proposals on inspection of earlier manuscripts and text versions in order to establish original, 'correct' spellings. It seems that there was some consensus about the spelling of individual words before Teika, at least to some extent on an etymological basis, as shown for example in the surviving handwritten letters of the wife of Fujiwara no Tamefusa (1049-1115) in which the distinctions between 1-p-1 and /-w-/ and between /.wo/ and /.o/ are maintained orthographically although the sound changes in (4) had already taken place, but it was Teika who explicitly established the etymological spelling principle which, together with his specific spellings of individual words, came to be known as Teika kana-zukai 'Teika kana-usage, Teika spelling' and soon acquired the status of a spelling norm, not just for copying, but also for composing text. Thus, from Kamakura to early Edo most literary writing was spelled according to the Teika kana-zukai. There are inevitably not a few mistakes in Teika's etymological spellings (for example too- 'far' < mid EMJ towo < OJ topo- spelled .!::: :a:- 'to.wo' instead of etymologically correct .!::: ~~ 'to.po'), first of all because the copies of texts he had access to and worked from already contained errors. Furthermore, his recommendations concern only a relatively small number of words and only address the use of ;:{_2,-r-..v \(Q V.:t3::a:- .2 Therefore later scholars elaborated on and amplified Teika's work. The Kanamoji-zukai (completed after 1363) by the monk Gy6a (fourteenth century, dates unknown, born Minamoto no Tomoyuki) is the first large-scale application of Teika's principles. It lists spellings for more than a thousand words and for a long time served as the major normative spelling reference. In addition to unsuccessfully attempting to credit his own grandfather, rather than Teika, with the establishment of the etymological spelling principle, Gyoa takes the important step of incorporating also the letter categories f~:bf'J: 5 b (as well as U which was 2

:a-,

However, with regard to :to and Teika partly employed a phonographic, rather than etymological, spelling principle, following the example of earlier dictionaries likelrohajiruisho (late Heian) in assigning letters in word initial position according to phonetic pitch, such that /.o-/ would be written with if it was [high], e.g. oku (
:a-

:a-<

6.1 Writing

175

problematic for entirely different reasons). Mention must also be made of the kokugaku scholar KeichU (1640-1701) who provided the basis for the rekishiteki kana-zukai in use today. In his Waji shoransho (1695) he set the etymological spelling principle on a firm philological footing with systematic citation for specific spellings of sources which even by today's standards mostly are reliable. 6.1. 3.1 Undoing the etymological kana-spelling In printed editions premodern texts are usually presented in the etymological kana-spelling, so that for example the word for 'front' is written "i"" regardless of the date of the text. Table 6.3 is a simple guide to phonemic transcription of EMJ and LMJ texts, giving the phonemic shapes represented by the kana letters ;:ti9%::a:-f'J:"(}b.r-...l~ in different periods, reflecting the sound changes which took place during the EMJ and LMJ periods (see 7.3, 11.2, 11.3). As these sound changes generally are not systematically reflected in the copies available to us today of most of the literary texts (due to the application of the etymological spelling principle in copying the texts), the assignment of texts to definite stages within these sound changes is to some extent arbitrary. This holds particularly for the texts dating just after 1000, including important literary texts such as the Genji monogatari, c. 1001-1010, and the Makura no soshi, c. 1000. It is clear that these texts postdate the change of /-p-/ to 1-w-1 (cf. 7.3.1) and that what is written ir>f'J: in editions of these texts represents /kawal. However, it is difficult to say for certain that the change of /wo/ to lol (cf. 7.3.2.3), which is thought to have been complete c. 1000, was in fact entirely complete by the time these texts were written, and it is therefore less certain that what is written ir·f~ in editions of these texts represents /kao/ and not /kawo/; it is quite certain, though, that it does not represent /kapo/. Note also that the difference between /.e/ and /.ye/, which pertains only to morpheme initial position, is not represented in the standard inventory of kana letter categories (cf. 6.1.2.4), but it is relevant to the earliest kana materials, such as Tosa nikki or Kokinwakashu; it can easily be verified in good dictionaries whether words written with initial ;:t had /.ye/ (e.g. yeda 'branch') or /.e/ (e.g. e- 'to get'); in morpheme non-initial position ;:t always represents /.ye/ before c. 950 (e.g. nuye 'thrush'). Through the EMJ and LMJ periods, G, i::J, ---::>, C, f, 16, ---5 straightforwardly represent /si, ti, tu, zi, zu, di, du/ and should be transcribed si, ti, tu, zi, zu, di, du, e.g. ---=>i::J tuti 'earth', G---5il• siduka 'quiet', J;.f mizu 'doesn't see', J;. C mizi 'won't see', and ~"i i6 yamadi 'mountain path'. Only in texts dating from the early NJ period do C and i6 both come to represent /zi/ and should be transcribed zi, and f and ---5 both come to represent /zu/ and should be transcribed zu. In terms of the etymological kana spelling (which in text editions supplies dakuten although these were not used in general writing, cf.

Table 6.3 Phonemic transcription of the historical kana-spelling < 950

c. 950

950 >

1000 >

1100 >

1300 >

Tosa nikki, Kokinwakashu, Ise monogatari

Taketori monogatari*

Kagero nikki, Utsuho monogatari, Ochikubo monogatari

Genji monogatari, Makura no soshi

Konjaku monogatari, Heike monogatari, Hojoki

Tsurezure-gusa, Soga monogatari

Kana

Examples

;t

;t

e

ilt.

yeda

eda

1b.;t

nuye

nue

b

wi

~

mawiru

VI::.&

wemi

~-

woka

oka

~

awo

ao

1:1&

para

tl>J-.t

kapa

b

.J. ~

1-.t

v-

~

pira kapi

.bJ:).

pune

ftls.

tapu

........

~

pera

l:f

*'l:ftJ•""

mape

tl>J:f

emi

kowe

tl•D-

.s.

i maim koe

fara kawa lira kawi

kai rune

tau fera mawe

mae foka

poka kapo

kawo

kao

Glosses: ;t 'to get', *.tC:. 'branch', 10-;t 'thrush',~ 'well', ;l:~.Q 'goeshwnbly' k:Ji- 'smile',:::.% 'voice', :§:f.p 'hill', 'other place', n>l::t 'face'. *Note that the dating of the Taketori monogatari is very tentative.

6.1 Writing

177

6.1.2.2), other kana letters than those discussed above have not changed in representational value since the beginning of the EMJ period and can be transcribed for premodern texts as they are for modem texts. 6.1.4

Sound tables

Today the kana letters are not presented in the Iroha sequence, but arranged in the gojiionzu, as shown in Table 6.1 above. This arrangement of the kana letters is now used universally in school education and for dictionary entries; it also influenced the katsuyokei system for describing verbal inflection (see 3.4.6). Kana lists (onzu {!(~ 'sound tables') which formed the basis of the current grid arrangement began appearing from the beginning of the eleventh century. The first known such onzu, which is incomplete consisting of only eight columns, appears at the back of the Kujakukyo-ongi (from around 1000) and the first complete onzu is found in Meigaku's Han 'on-saho (1093). As opposed to the Iroha, the onzu went beyond simply enumerating the distinct orthographic categories: the onzu involve a substantial phonographic and phonological analysis and systematization of the letter categories defined in the Iroha. They provide a phonological analysis into consonants and vowels of the moras represented by the kana letters, such that kana in a column share the same initial consonant and kana in a row share the same vowel, i.e. an analysis of a syllable into subsyllabic parts, or a mora into submoraic parts. In the early onzu the kana were not actually arranged in grids with columns next to each other, but as lists with the columns following each other, but the principle is the same, of columns sharing an initial consonant and the vowels appearing in a fixed order within the columns. The grid system, incidentally, also makes it possible to refer in Japanese to consonants, which have no separate kana representation, with reference to this arrangement, e.g. the 'consonant of the ka-column' (ka-gyo no shi 'in :JJ fi"0)-1-lf), viz. k. Ingenious as they were, the onzu remained in the realm of scholastic and academic writing and annotations until late in the Edo period; in the Meiji period the gojiionzu was adopted for educational purposes and eventually replaced the Iroha as the presentation of the kana categories and as an organizing principle for dictionaries. The first Japanese dictionary ordered in the gojiionzu sequence was the Genkai ( #JJ) published in 1891 by Otsuki Fumihiko, who was also an important figure in the development of a 'standard' Japanese, see 13.2.2. The phonological analysis of the onzu was principally inspired by Chinese rhyme tables and in particular by the fonqie (~ "!W) spelling system, in which the 'reading' of a kanji is described in terms of two other kanji, in a form like 'x yz' to be understood as 'x has the initial of y and the rhyme of z', for example, Jl:t i.i&IB: which means that Jl:t has the initial of fg (EMC *t;)k, LMC

s

178

6 Writing and sources

*t::1~k) and the rhyme of ;'f-I (EMC *y::JWIJ, LMC *x.fi::JWIJ), and thus that Jl:i: has the readings EMC *bwlJ, LMC *t::lWlJ. However, knowledge of Indic script, again in particular the Siddham script (cf. 6.1.2.2), exerted an important influence on the development of the sound tables. The sequence of vowels ( &> a, v' i, ? u, ;t e, .to o) as well as consonants (il" k-, ~ s-, tc t-, ft. n-, l'i p- (> f- >h-), ~ m-, ~ y-, 1::> r-, :b w-) in the gqjilonzu corresponds to that of the letters in the Siddham script. The letters representing true consonants come first, followed by approximants and the liquid: y, r, w. For the true consonants, this arrangement moves from velar to labial place of articulation, with nasals following oral consonants. The position of s- in the arrangement has been taken to support the view that early EMJ /s/ phonetically was not a simple sibilant (which in Siddham are placed at the end together with approximants and liquids), but rather an affricate (2.2.3); the arrangement also clearly shows that the reflex of OJ /p/ was an obstruent, most likely a stop, /p/, and certainly not an approximant such as bilabial [cp] (7.3.1.3). Between the earliest onzu from the beginning of the eleventh through the twelfth century, there was considerable variation in the order of vowels or consonants. The first onzu, from the Kujakukyo-ongi, has the vowels in the order i, o, a, e, u and the consonants k, s, t, y, m, p, w, r. The onzu in Meigaku's Han 'on-saho has the vowels in the current order and the consonants ordered strictly by place but not manner of articulation: k, y, s, t, n, r, p, m, w. Eventually the order we know today, which is first found in the Shittan-yojil-ki (1075) by Kanchi (1 045-1111 ), was settled upon. The sound changes mentioned in (4) above also affected the phonological analysis and arrangement of the kana in the onzu. In particular, when the kana pairs .to ::a:, v'?;;>,and ;t% became phonographically equivalent, standing for /.o, .i, .e/, respectively, the historically correct assignment of these letters to the a-gyo or wa-gyo within the onzu became far from obvious, and until quite late there are many examples of onzu with incorrect assignments. In his 1682 Shittansanmitsusho, the monk Jogon finally established v' and d;;> in the a-gyo and wa-gyo, respectively, and KeichU (1640-1701) did the same for ;t and 2. in 1693 in the Wajinshoransho. It was not until 1776 that .to and were finally fixed in the a-gyo and wa-gyo, respectively, in Moto'ori Norinaga's Jion kana-zukai.

:a:

6.2

Sources

The sources available to study EMJ are far more extensive and varied than the sources for OJ. In particular, we have from the middle of the period a large set of prose texts, written in kana in a language form which is thought to be close to the vernacular of the nobility and officials at court and which gives a more comprehensive picture of the contemporary language than we have for any other time in premodern Japanese. Sources are scarce until the early tenth

6.2 Sources

179

century, and there is thus a substantial gap of more than one hundred years in the solid attestation of Japanese. This is usually attributed to the fact that written Chinese almost succeeded in establishing itself as the written language in Japan at the expense of Japanese, particularly outside poetry, but thankfully, from a linguistic and cultural point of view, that did not happen. Apart from the first one hundred years, the He ian period is characterized by a lively culture of writing in Japanese which produced some of the masterpieces of Japanese literature, such as the Genji monogatari and the Makura no soshi. After around 1100 the language in the written sources increasingly fossilized and towards the end of the Heian period the written norm known as 'Classical Japanese' (bungo X~ 'written, literary language') became relatively fixed and served as the dominant base for writing in Japanese until the beginning of the twentieth century. The sources for this period, in addition to the prose and poetry texts (6.2.1), include a body of annotated Chinese texts, the so-called kunten shiryo (6.2.2), as well as dictionaries (6.2.3) and Sanskrit-studies (6.2.4). Table 6.4 is a chronological list of some of the important sources. 6.2.1

Prose and poetry

Prose texts are by far the most valuable material for studying the language of this period. There exist anum her of poetry anthologies and many poems within prose texts from the He ian period, but while the literary merit of some of these poems is considered high, their value for linguistic studies is more limited: a norm concerning grammar and lexis soon became established in poetry, and most of the poetic texts are therefore conservative in these regards, for example avoiding both SJ vocabulary and native forms with bound moras (in particular onbin forms). In this respect the Heian period poetry is in stark contrast with the poetry from the Nara period, which constituted the major source of OJ. Among the poetry anthologies from the period, the first one, Kokin wakashil (i!4-f1Jm?\. 'Collection of Japanese poetry'), stands out, but in particular for its Japanese preface (kanajo -f&:tl.Ff. 'kana-preface') on poetics, written in hiragana by the editor Kino Tsurayuki (?868-?945). This is the first piece of prose writing in hiragana and it contrasts with most later hiragana writing from the period by being expository and non-fictional. The prose material includes monogatari ('stories') of various sorts, ranging from collections of short stories, or even what would today be called 'shortshort stories', e.g. Ise monogatari ({¥~~'Tales from Ise'), to Murasaki Shikibu's very long novel Genji monogatari (i!J:.B:;;~~ 'Tale of Genji ')~ other well-known works are Taketori monogatari (tt:®.~~ 'The tale of the bamboocutter')/ Utsuho monogatari (~$*~~ 'The tale of the hollow tree'), 3

Taketori monogatari is thought to date from the middle of the tenth century, but the earliest extant copies date from the Edo period, making it less reliable as linguistic evidence than the other EMJ texts.

180

6 Writing and sources Table 6.4 Important EM.! sources Ninth century Konkomyo saishoo kyo (annotations c. 830) Todaiji fujumonko (before 834) Zaitoki (c. 842) Shinsenjikyo (898-901) Tenth century Knkin wakashii (914) Ise monogatari (early tenth century) WamyO-ruiju-sho (c. 934) Tosa nikki (935) Taketori monogatari (mid tenth century) Kagero nikki (second half of tenth century) Utsuho monogatari (970s) Ochikubo monogatari (late tenth century) Sanbo ekotoba (984) Eleventh century Hokke mongu (annotations c. 1000) Makura no soshi (c. 1000) Genji monogatari (1001-10) Tsutsumi chiinagon monogatari (c. 1055) Sarashina nikki (1059-60) Konkomyo saishoo kyo ongi (c. 1079) Han'on saho (1093) Ruiju-myogi-sho (c. 1100) 1\velfth century Daijionji sanzohoshi-den (annotations 1099-1116) Konjaku monogatari-shii (c. 1120) Shittan yoketsu ( 11 01) Okagami (c. 1119) Irohajiruisho (compiled between 1144 and 1181) Knhon setsuwa-shii ((late?) twelfth century) Shittan kuden (c. 1180)

Ochikubo monogatari (1t~~Mf 'The tale ofOchikubo'), Tsutsumi chilnagon monogatari (~9l*fi~~Mf 'The tales of the Tsutsumi middle counsellor'), and the historical novel Okagami (:lc~ 'The great mirror'). Other genres include nikki ('diaries'), which mostly were not private, but written for circulation, such as Ki no Tsurayuki's Tosa nikki (±{;£ 13 We 'The Tosa diary'), the Kagero nikki (~W~~ 13 We 'Gossamer diaries') or the Sarashina nikki C~~ 13 WC 'The Sarashina diary'); and zuihitsu (~* 'jottings'), especially Sei Sh6nagon's famous Makura no soshi (tt1it-1- 'The pillow book'). Much of this literature was written in hi ragana and by women in a free-flowing, lively, elaborate and sometimes playful language.

6.2 Sources

181

The so-called setsuwa (~~'tales, legends') literature comprises a variety of short tales ranging from didactic Buddhist stories set in India or China to local folktales and ghost stories. The setsuwa literature is generally written in kanji-kana majiribun (6.1.1), i.e. with a large proportion of kanji. The language in the setsuwa and other writing in kanji-kana majiribun is relatively simple and straightforward compared to the hiragana literature, but it is also more heavily and obviously influenced by Chinese, because of the influence of kanbun-kundoku (see 9.1) on kanji-kana majiribun. The most famous anthology is the Konjaku monogatari-shii (4-tf~~· 'A collection of tales of times now past'); others from this period are the Kohan setsuwa-shii (if:;ljs: ~~!·'A collection of old tales') and Sanbo ekotoba (-=.3s:~~ij] 'Illustrated stories about the three jewels'). As the written language fossilized, the narrative was kept in the classical written language, but dialogue was in a fonn close to the vernacular. From the end of the period, and especially in the LMJ period, the dialogue passages in the setsuwa literature are important sources of changes in the language, as most other writing was done in the by then fossilized classical written language. These types of prose texts together present a comprehensive view of the contemporary language and its use. As with materials from the Old Japanese period, the text portions written logographically are less useful for studying the lexicon and grammar of the language than those written in kana. However, because of the volume of text this is in practice not a severe problem for studying the grammar. The texts from this period have come down to us in later copies, subject to scribal error and to conservative and nonnative redaction, especially in applying the etymological spelling principle (6.1.3), making them unreliable as guides to the dating of sound changes. However, much effort has been invested in producing reliable critical editions of the texts. Most editions present the texts in etymological spelling and, for the texts written in kana, with many interpolated kanji (for the convenience of modem readers), but good editions will also indicate the written fonn of the base text. 6.2.2

Annotated texts

Kunten shiryo (~Jilg~;j!.l} 'materials with reading marks') are texts in Classical Chinese with annotations which provide a guide to the pronunciation of Chinese or SJ words, or to understanding or rendering the texts into Japanese. See below (9.1) for details about kunten, kuntengo (the language used in kunten texts), and kanbun-kundoku, the practice of rendering Chinese into Japanese. A great amount of kunten texts exist, mostly unpublished and in the possession of Buddhist temples. A great effort of both fieldwork and philology has been and is being invested in gaining access to, charting, cataloguing and studying these materials. Impressive advances have been made in

182

6 Writing and sources

recent years, but there is still enonnous scope for further research, in addition to the sheer volume of material also because the material is complex and difficult to interpret. The study of kunten texts and kuntengo has become a major focus for research for Japanese linguists and philologists since the 1950s, more recently attracting overseas scholars as well. Kunten texts are important, primary sources. Often the date of annotation and the identity of the annotator are noted in the text. It is in particular thought that early kunten texts from the late eighth and ninth centuries reflect some fonn of contemporary vernacular language and that they thus can contribute to filling the gap left by the lack of other types of sources between late OJ and 900. Glosses provide valuable infonnation about vocabulary not attested elsewhere and often provide an accurate indication of pronunciation and therefore of contemporary phonology, making them important material for dating sound changes, as opposed to the versions of the prose texts which we have today. It must be kept in mind, however, that kunten texts are annotations and therefore orthographically severely underspecified and on some points very difficult to interpret, so their evidential value is not straightforward. In addition, strict, dogmatic nonns of rendition and annotation arose, so that kunten texts from after the middle of the EMJ period generally cannot be taken to reflect contemporary language, but especially in grammar and vocabulary had become conservative and even archaic. As mentioned above (6.1.2), the kunten material is also very important to the study of the history of the development of the kana scripts. The amount of published studies and material increases steadily, but well-publicized kunten texts include Konkomyo saishoo kyo (~Jiti~IH!fdl~::E{€£ 'Sutra of the golden light and most victorious king', annotated c. 830), the Hokke mongu U~~Jt 10 'The sentences and phrases in the Lotus sutra', c. 1000, see 9.1.1 ), and the Daijionji sanzohoshi-den C*~~~~~¥~Gfli~, Kofukuji manuscript, annotated 1099 and 1116).

6.2.3

Glossaries and dictionaries

Closely related to the kunten materials are so-called ongi ('glossaries', if~ 'sound-meaning') which are lists of words extracted from individual texts in Classical Chinese, explaining the pronunciation or meaning of words or characters and aiding the reading of individual texts. There are a num her of ongi from the period. An important example is the Konkomyo saishoo kyo ongi (Glossary for the 'Sutra of the golden light and most victorious king', 1079) which also contains the first attestation of the gojilonzu (6.1.4) and of the iroha-uta (6.1.2.4). Originating as combinations of glossaries for single texts, the period also saw the compilation of the first dictionaries, which were independent of individual texts. These include the Shinsen jikyo Oi~lr~.l*~), Wamyo-ruiju-sho (~45~~:f) or Wamyosho ~45 :f)), Ruiju-myogi-sho (~

6.2 Sources

183

~45~F)), and the Irohajiruisho (~~*~¥)or {¥g;m:*~F)), compiled in the second half of the twelfth century and the first dictionary ordered according to the sequence of letters in the Iroha-uta. Meigaku's Han 'on saho (.&if ff¥~) explains the pronunciation of kanji by thefonqie spelling principle (and by kana glosses).

6.2.4

Sanskrit studies

The term shittangaku (~ft~), which literally means the study of the Indic script Siddham, is used to refer to the study of Sanskrit and materials in Sanskrit related to Buddhism. It comprises a number of materials which are important to the study of Japanese, especially its phonetics, because they explain or exemplify features of Sanskrit by comparison with Japanese, although the interpretation of the material in many cases is far from straightforward. They include the Zaitoki (1":EJi!f~ 'Record of a sojourn in China'), the priest Ennin's notes on the pronunciation of Sanskrit (which he studied in China) illustrated by means of sinograms used as phonetic (rather than phonemic) symbols, based on their sound values as man 'yogana or their Japano-Chinese readings (see 9.2.1 ), supplemented with comments; Meigaku's Shittan yoketsu (~ft~~ 'Essentials of Sanskrit learning') which has many valuable observations on Japanese in order to illustrate points about Sanskrit pronunciation; and Shimen's Shittan kuden (~ft 0 fzi 'The secrets of Sanskrit learning') which exemplifies Sanskrit sounds by notes and by Japanese sounds. REFERENCES Writing, general: Kono 1969, Mabuchi 1993, Olmo 1977, Seeley 1991, Tsukishima 1977: 52-120, Tsukishima 1986. Runes: Elliott 1959: 14ff, 21ff. Hentaigana for 1-t!: Sugahara 2000: 23ff. Sources: Tsukishima 1987: 12ff Kunten materials: Kasuga 1985, Nakada 1969, Nishizaki 1992, Tsukishima 1986, Yoshida et a/. 2001. Dictionaries: Nishizaki 1995.

7

Phonology

The major phonological changes which distinguish OJ from NJ took place through this period, leaving Japanese at the end of the period phonologically looking much like NJ. The main structural change is the introduction of the distinction between short and long syllables and the emergence of bound moraic phonemes which occurred after a free mora to form a long syllable. At the end of the period, EMJ had the sound inventory shown in Table 7.1. The phoneme inventory in free moras was the same as in OJ, but a number of changes had taken place concerning their distribution. Of the free moras in Table 7.1, /wi, we/ were almost exclusively found only in word initial position (cf. 7.3.2.3), and /pa, pi, pu, pe, pol in word initial position and after /Q/ (cf. 7.3.1; also 11.3), whereas /b, d, g, z, r/ as in OJ were not used in word initial position, except in SJ loanwords. The intake of SJ loanwords also introduced syllables with complex onsets /Cy-/ and /Cw-1. The traditional term for these moras is yo'on 'twisted sounds' WJ{f which is opposed to choku'on 'straight sounds' ill[{if, i.e. moras with simple onsets. Of the bound moraic phonemes, 1-t/ was used only in SJ vocabulary, which is also often said to have had a distinction between syllable final-nand -m. See further 7.2 about some of the special phonological features of SJ vocabulary. Whereas OJ writing seems to have represented the contemporary phonology well, EMJ writing was, as explained above, underspecifying in several respects, no longer noting the distinction between tenues and mediae in general writing (6.1.2.2), and having no distinct ready means of writing new phonemes in the language (6.1.2.5). For that reason, the phonemics and phonetics of the new moraic phonemes are not directly or systematically reflected in the written sources. This holds in particular for the feature of nasality. However, interpreting the sources in the light of what we know or can reconstruct of OJ phonology and the sound changes involved in the introduction of the new phonemes (7.1.4), combined with later reflexes and writings of the forms in the texts, we are able to obtain a fairly detailed picture of the phonological system of the moraic phonemes also during EMJ. The late LMJ Christian materials provide an invaluable reference point for consonantal nasality, but the distinction we posit between 11, U/ and li, U/ had been lost by late LMJ. As we saw in 6.1.2.2, 184

7.1 Syllable structure

185

Table 7.1 EM/ sound inventory Free moras a ka ki u ku e ke 0 ko

sa si su se so

ta ti tu te to

ga gi gu ge go

za zi zu ze zo

da di du de do

kya (kyu) (kyo)

sya syu syo

tya (tyu) (tyo)

gya (gyu) (gyo)

zya zyu zyo

dya (dyu) (dyo)

na ni nu ne no

pa pi pu pe po

rna mi mu me mo

ya yu yo

ra ri ru re ro

wa wi we

ba bi bu be bo nya (nyu) (nyo)

pya (pyu) (pyo)

mya (myu) (myo)

rya (ryu) (ryo)

bya (byu) (byo)

kwa kwe gwa gwe Bound moraic phonemes Vowels /-1, -U, -1, -0/ Consonants 1-N, -Q, -C, -t/

some pronunciation glosses to Chinese texts note [i, fi] as distinct from [i, u], but the main reason we reconstruct this distinction for Japanese in HMJ is the observable traces in morphophonological rules or in lexicalizations of the postnasal neutralization rule, whereby tenues became mediae after nasals (7.1.2.2). Concretely, we posit Ji, U/ in onbin (and related) forms (7.1.4.3) where the source syllable had initial /b, g, m/, and in SJ vocabulary which reflects EMC *-IJ.

7.1

Syllable structure

EMJ and later stages of Japanese are set off from OJ by a major change in phonological structure: the introduction of the distinction between metrically shorl (or light) and long (or heavy) syllables. Through this change Japanese became quantity sensitive and the mora became a relevant unit in the language. Moras are measures of syllable length (sometimes more concretely thought of

186

7 EMJ phonology

as syllable constituents) of which a short syllable consists of one and a long syllable of two. Whereas OJ only had syllables with a single vowel as peak and nucleus (2.5), EMJ acquired long syllables of the form evv and eve, which have an additional position or constituent in the syllable after the peak vowel; see 7.2 about complex onsets (e.g. Cy- or ew-) which do not contribute to syllable length. The structure of the syllables of EMJ may informally be illustrated as in (1), which simply has long syllables with branching rhymes, with a new syllable position within the rhyme. (Depending on the theoretical framework of phonological representation, long syllables are usually more formally represented with a complex (or branching) syllable nucleus or a syllable coda.) (1) a. tauto- 'exalted' Long evv sy liable

Short syllable

cr

cr

~

0

R

A

0

A

R

e

v

v

e

v

t

a

u

t

0

b. punde 'brush' Long eve syllable

Short syllable

cr

cr

~

R

0

0

A

R

e

v

A

e

e

v

p

u

n

d

e

7.1 Syllable structure

7.1.1

187

Bound moraic segments

This general syllable structure, with a position after the peak, has remained part of the phonological make-up of Japanese since EMJ. The phonemes which occur in this 'post-peak' position in the syllable are moraic, because they measure, or constitute, one mora, and they are bound (or dependent) because they do not form a syllable peak, but occur after the syllable peak. The EMJ system of bound moraic phonemes is somewhat different from NJ which has II, U, V, Q, N/. The overall EMJ inventory was as shown in (2); these phonemes are customarily noted in capital letters. (2)

Vowel Front Back

+

Nasal

II

i

+

u u

Consonant

+ a. Q N C/

II, i, U, U/ are high vowels, distinguishing front versus back, and oral versus nasal; /Q, Nl are consonants whose only distinctive feature is nasality and which are not otherwise specified for place and manner of articulation, but copy these features from the following consonant, as in NJ; and /C/ is a consonant which is not even specified for nasality, but copies all features, including nasality, from the following consonant, see ( 4) in 7.1.2.1. /Q/ was only used before /p, t, k, s/, and /C/ was not used before /r, w, y/. Other than that, the distribution of these segments depended on position in the word and morpheme, (3). It is sometimes said that two moraic nasal consonants were distinguished, IN/ and /M/, but see below (7.2). (3)

morpheme non-final word final morpheme final, word internal

II, U, C/ II, i, U, U, N/ II, i, U, U, Q, Nl

It should be noted that few descriptions recognize the nasality distinction among the vowels, and most assume that nasality was phonemic for the consonants in all positions. That is to say, most descriptions posit a system for EMJ which is identical to that for NJ, and we will therefore in the following discuss the system we here posit for EMJ. Recall that the written representation of the new bound m oraic phonemes was incomplete and inconsistent (6. 1. 2. 5. 1) and that the difference between tenues and mediae was not noted in general writing (6.1.2.1). This means that the phonemics and phonetics of the new moraic phonemes are not directly or systematically reflected in the written sources. What we can say is based on kunten glosses (the interpretation of which is not always straightforward) and what we know or reconstruct from OJ phonology and from the sound changes involved in the introduction of the

188

7 EMJ phonology

new phonemes (7.1.4), combined with later reflexes and writings of the forms in question (see 11.1).

7.1.2

Nasality

The major difference between the EMJ system and NJ lies in the distribution and distinctiveness of the feature of nasality in the bound moraic segments. In EMJ nasality was distinctive for both vowels and consonants (/I, U, Q/ versus li, 0, N/) in morpheme final position, but not distinctive for either consonants or vowels in morpheme non-final position. In the inventory of bound moraic phonemes found from LMJ onwards (/1, U, V, Q, N/ cf. 11.1.2), on the other hand, bound moraic consonants have distinctive nasality, whereas moraic vowels do not, in both cases regardless of position in the morpheme. This difference is closely linked to the phonetic prenasalization of the mediae (2.2.2) and the overall role of the feature of nasality in OJ and MJ.

7.1.2.1 Morpheme internal position In morpheme internal position, moraic consonants in EMJ only occurred before /p, t, k, s, b, d, g, z, m, n/. What we have reflected as NJ /Q/ and IN/ in morpheme internal position were in EMJ in complementary distribution, with phonetically oral moraic consonants occurring only before /p, t, k, s/, e.g. tatto- 'precious', and phonetically nasal moraic consonants occurring only before /b, d, g, z, m, n/, e.g. kanbasi 'fragrant', punde 'brush' (> NJ hude), pingasi 'east'(> NJ higashi), wonna 'woman'(> late EMJ anna). There were in other words no minimal pairs such as NJ sittai (/siQtai/) 'blunder' versus sintai (/siNtai/) 'body'. Consequently, there was no distinction between IN/ and /Q/ in morpheme internal position and nasality was not distinctive in the moraic segments in that position. In other words, in morpheme internal position there was only one moraic consonant, /C/, which was phonemically unspecified for nasality, but derived its phonetic nasality value from the following consonant. /C/ was [-nasal] before tenues (/p, t, k, s/), but [+nasal] before nasals and mediae (/m, n, b, d, g, zl). This makes it easy to understand the overlap in the writing of oral and nasal moraic consonants in early materials mentioned in (6.1.2.5.1), as they in morpheme internal position were not distinct phonemes, but conditioned automatic variants of one phoneme /C/. The forms just cited were therefore phonemically /taC.ta, kaC.ba.si, puC.de, piC.ga.si, woC.na/, with'.' showing a syllable boundary. The phonetic nasality value of /C/ was copied from the following consonant (of which lb, d, g, zl were phonetically prenasalized: ['1>, nd, ~g, nz]), see (4), alongside other features of manner and place of articulation. ( 4)

/C/ => [a.nasal] I_ [a.nasal]

7.1 Syllable structure

189

This is a simple phonetic rule, anticipatory assimilation of the phonetic nasality of the following nasal or media, or, leftwards spreading of nasality, in the same way that vowels preceding nasals and mediae were nasalized (2.3). This is shown in (5) in a phonetic notation which shows first prenasalization (and medial voicing oftenues) and then nasality spreading: (5) /kaC basi/ /puC de/ /piC ga sil /woe na/

Prenasalization => [kaC mba 3i] => [puC ndie] => [piC Dga 3i]

Nasality spreading => [kamba3i] => [piindie] => [pii)ga3i] => [won:a]

7.1.2.2 Morpheme final position; postnasal neutralization On the other hand, nasality seems to have been distinctive both for moraic consonants and vowels in morpheme final position. The main reason for positing this distinction is the effect it had on a following tenuis. This is mainly relevant in verb inflection and in SJ loanwords. For example in the innovative EMJ gerund forms in (6a) (see 8.1.4) the most straightforward interpretation is that they derive from underlying phonemic forms as in (6b). As opposed to monomorphemic forms such as punde /puC.de/ mentioned above, where /d/ was prenasalized [nd] and provided the phonetic nasality, the nasality in these forms must have derived from the moraic segment. (6)

a. sin-de 'dying' yofi-de 'calling'

<= <=

b. /siN + tel ( <= sin-) /yoU+ tel(<= yob-)

sit-te 'knowing' ou-te 'pursuing'

<= <=

/siQ +tel ( <= sir-) /oU +tel(<= op-)

The rule which derives the (a) forms in (a) from their underlying (b) forms is a neutralization rule such as (7) which neutralizes the distinction between tenues and mediae after nasal segments. This rule was not limited to the formation of verb forms, but applied to all combinations of underlying morpheme final li, 0, N/ + /p, t, k, s/, including in SJ vocabulary, where we find many words which have lexicalized the effects of postnasal neutralization, for example safizi(n)-mono 'vegetarian food' (< safi+sin *i~ 'devotion to Buddhism; abstention' + mono 'thing, stuff'), safizok- 'dress up' < safizoku 'dress'< safi+soku; see further (11.1.1). Interestingly, postnasal neutralization was, at least in LMJ, the subject of an explicit reading rule, umu no sita nigoru (see 11.1.1.1). (7)

/p, t, k, s/ => lb, d, g, zl I II, 0, N/ _

190

7 EMJ phonology

Postnasal neutralization suggests strongly that medial voicing of the tenues took place also after IN/, which, combined with prenasalization of the mediae, made it close to impossible tom aintain a distinction between tenues and mediae after nasals: /siN/+ /tel=> [sin]+ [dje] versus /siN/+ /de/=> [sin]+ [ndje].

7.1.2.3 Word .final position In word final position, /Q, Cl were not used. II, U/ were used in both native and SJ vocabulary. li, 0, N/ were mostly used in SJ vocabulary where they were frequent, but at least IN/ and /U/ were also found in some native forms, such as -N or - 0' (reduced variants of -mu, the conclusive/adnominal of the conjectural -(a)m-, cf. 8.5). In LMJ, word final IN/ had the sound value [n] (see 11.4.2). We have little positive basis for determining the EMJ sound value, but it seems likely that it was [n]. 7.1.2.4 Nasality harmony Although the two rules of nasality spreading (4) and neutralization after nasals, (7), are different types of rules, their effect was much the same. In both cases the rules result in a certain amount of nasality harmony. 1 Compare two sets of 1

In Frellesvig (1995), I suggested that OJ and MJ had a process of ambisyllabication which consists in linking up a non-final syllable with a following syllable, through an underlyingly empty coda slot. On that analysis, the domain of nasality harmony is the syllable rhyme, whose nasality could derive from the onset consonant of the following syllable, incorporated as coda through ambisyllabication, as in [pOn~e] <= /puCde/:

a

(i)

~

~

R

0

0

R

1\

N

Coda

~-------------c

v

c

c

v

e

p

[+nasal][-nasal]

7.1 Syllable structure

191

homophonous words: [pilndie]: (a) a monomorphemic noun 'brush' and (b) a verbal gerund 'stepping'; and [tiii.ndie]: (a) a monomorphemic noun 'order, sequence' and (b) a verbal gerund 'following'. Although phonetically identical, we posit different underlying phonemic forms and different derivations, see (8). Note that in the (b) forms, prenasalization had little effect and that nasality spreading took place from the nasal moraic segment to the preceding peak vowel.

(8)

a. /puCde/ 'brush' /tulde/ 'sequence'

Pre nasalization => [puC ndie] => [tui ndie]

Nasality spreading => [pilndie] => [tiii. ndie]

b.

Postnasal neutralization /puN+ tel 'stepping' => /puN de/ /tul + tel 'following' => /tul de/

Nasality spreading => [pilndie] => [tiii. ndie]

Nasality harmony was an important feature of OJ and HMJ phonology, together with prenasalization of the mediae making nasality a prominent feature in the phonetic make-up of OJ and EMJ. Nasality, postnasal neutralization, and medial voicing made OJ and EMJ sound very different from late LMJ and NJ, in which prenasalization, nasality harmony, postnasal neutralization, and medial voicing were all lost, cf. 11.1.

7.1.3

Transcription ofmoraic segments

The notation ofmoraic segments by capital letters looks clumsy. Furthermore, the morphological constituency of a form is not always obvious, making the notation or not of phonemic nasality arbitrary, so unless it is relevant in the context and we wish to show this in phonemic transcription, we shall use a semi-allophonic notation and transcribe forms such as /taCto-/ 'precious' as tatto-, /siQ-te/ as sitte, /puCde/ and /puN-te/ as punde, and /tulde/ and /tul-te/ as tuide.

7.1.4

Sources of long syllables: the onbin sound changes

The phonological changes that triggered the introduction of long syllables were the so-called 'onbin' sound changes, which occurred some time between the late eighth and the early tenth century. The term onbin (iff~) 'euphony' is used in Japanese philology to refer to a variety of phenomena, including both sound changes and morphophonological processes, which have in common that they are conceived of as being motivated by ease of articulation

192

7 EMJ phonology

and involve the change of a CV syllable to a bound moraic segment. (Earlier, onbin was used in a wider sense to include also rendaku (2.6.2) or renjo (7.2.1, 11.4.21.) It is no exaggeration to say that the onbin changes comprise the defining set of linguistic changes of the EMJ period, which left the language looking at the end of the period much like NJ in phonology and morphophonology. First of all, the onbin changes led to the introduction of the new syllable structure and system of bound moraic phonemes described above, resulting in a major typological change in the phonological structure of Japanese. Second, they resulted in changes in the morphophonology of verb and adjective inflection, as the sound change affected the shape of inflected forms and eventually was regularized to give consonant base verbs a new stem derived from the infinitive (e.g. OJ ipi- 'say' > EMJ ipi- - iU-IiQ- or yomi- 'read' > yomi- yoU-lyoN-, see 8.1.4) and the adjectives new shapes of some forms (e.g. taka-ki 'tall-ACOP.ADN' > taka-! and taka-ku 'tall-ACOP.INF' > taka-U); see 8.2). Table 7.2 gives some examples. Those listed under (a) exemplify the core sound changes involved in onbin (see 7.1.4.1); those under (b) are morphophonemic changes analogically modelled on the sound changes, and (c) are sporadic word individual changes which fall outside the core sound change, but which yielded forms with one of the new phonemes. The outcomes are listed in two columns, depending on whether the resulting bound moraic phoneme was a consonant or a vowel. In addition to their far-reaching influence on the phonology of Japanese, there are several noteworthy features which set the onbin changes off from regular sound changes. First, the onbin changes give rise to variant shapes of individual words or morphemes which coexisted for a considerable period. Second, they were non-automatic sound changes; that is to say, there were certain phonological conditions which had to be met for onbin to take place, but no identifiable phonological conditions under which the changes must occur. Third, as shown by the examples in Table 7.2, some source syllables gave variable outcome, either a moraic consonant or vowel. Onbin forms appear in the written sources from the beginning of the EMJ period. These forms are first of all found in Buddhist kunten glosses, which to a large extent reflect the contemporary vernacular (see 6.2.2). A relative dating based on phonological criteria shows first that the onbin changes predate the merger of intervocalic /-p-/ with /-w-/ which took place in the second half of the tenth century (7.3.1.1): (a) 1-pV!, but not 1-wV! syllables gave onbin; (b) onbin took place before /-p/, but not before /-w/. Furthermore, OJ !Cil syllables took part in the onbin changes, whereas OJ /Cwil syllables did not. OJ /Cwi! was much less frequent in the lexicon than /Ci/ and this correlation could therefore be fortuitous, but it is significant that UB verbs whose base

7.1 Syllable structure

193

Table 7.2 Examples of onbin Vocalic

Consonantal

niuta omoute

nitta omotte

laputwo- 'exalted'>

laulo-

latto-

ald + bilo 'trade + man' > ywobite 'caii.GER'

aldiido yoiide

aldndo yonde

(a) Core sound changes nipi + Ia 'new + field; proper name' > omopite 'think.GER'

*sabusabu-si 'lonely'>

saiizail{-si)-

kwo + miti 'small + road' > lumite 'pluck.GER'

koiidi luiide

kondi Iunde

pi + muka(si) 'sun-facing, east'>

piiiga 'proper name'

pingasi 'east'

salci-dat- 'ahead-set.out, precede'> okite 'put.GER' > yo-ki 'good-ACOP.ADN' >

saidatoite yo-i

kani-pata 'crab-loom, silk brocade'> sinite 'die.GER' >

kanbata sinde

Jdnu + kaJd 'silk+ fence; silk curtain'

kingai

lakume 'wholly' > yo-ku 'good-ACOP.INF'

laume yo-u

lugite 'order, sequence' > kogile 'row.GER'

luide /wide

kagupasi- 'fragrant'>

kailbasi-

(b) Purely morphological onbin motile 'hold.GER' lotite 'take.GER' idasile 'take.out.GER' aru meri 'exist EVID'

(c) Sporadic word individual changes mawos- 'say.HUM' mawide 'visit.HUM' *kari-ta 'harvest-field; proper name' *kari-da 'harvest-field; proper name' kagapuri 'headwear (on the head)' kadilori 'oarsman'

kanbasimotte lotte

idaite anmeri mausmaude 'shrine/temple visit' kaJta kanda kailburi kandori

and infinitive in OJ ended in /-wi/ did not acquire onbin stems (8.1.4), see (9). If the onbin changes had taken place after the merger of /Ci/ and /Cwil there is no good phonological reason that the UB verbs should not have acquired onbin variants. This means that the onbin changes in the main predate the merger of /Ci/ and /Cwi/ (7.3.2.1), i.e. they must originate in the OJ period.

194

7 EMJ phonology

(9)

QD UB

'put' 'arise'

OJ okiokwi-

EMJ oki- ~ oioki-

Thus although both the kunten materials and a relative dating of onbin show that onbin fonns were an established part of the language by the time of the appearance of the first mainstream EMJ written sources at the beginning of the tenth century, onbin forms are underrepresented in the literary sources and remained so through the entire MJ period. In particular, onbin fonns are virtually never found in poetry, because of the conservative nonn in poetic language (6.2.1). 2 On the whole, we are not able to say much in detail about the course of use and spread of onbin forms through MJ on the basis of the written sources, for onbin forms never became part of the written nonn which emerged in the course of the second half of the EMJ period. However, as may be expected we find onbin fonns far more widely represented in prose writing, but also there the impression is that onbin fonns are often used to impart a flavour of spoken language. This is confinned by remarks by the priest and Sanskrit philologist Meigaku in his Shittan yoketsu (11 01) where he characterizes onbin fonns as being characteristic of spoken language, but the source forms as being used in writing (cited in Mabuchi 1971: 92). His examples are mainly various fonns of morphological onbin, e.g. be-i NEC-ACOP.ADN being used in speech instead of be-ki which was used in writing and therefore considered more correct, or tot-te 'take-GER' being used instead of tori-te, but he also gives lexical examples, saying for example that komiti (< OJ kwo-miti) is the correct fonn, but that people popularly say koiidi. These remarks, some centuries after the onbin sound changes took place, reflect both a tension between written conservative nonns and spoken language, as well as the fact that the onbin changes first of allied to the establishment of variant fonns of words or morphemes, in many cases not the replacement of an older form by a newer. The variation that was established was in some cases stylistic (like the variation in NJ between watakushi, watashi, atashi for T), and in others it became grammatically conditioned, as with the variation between infinitive and onbin stem of consonant base verbs, e.g. OJ sini- 'die.INF' > EMJ sini(INF) ~sin- (onbin-stem) subject to selection by different contexts (8.1.4). In some cases the onbin fonn eventually replaced the source form, e.g. kaki-mami- ('peek at through a fence; lit. fence-space-look') > kaimami- 'have a glimpse of', or the adjectival adnominal fonnant -ki > -i. Conversely, in other cases it was the source fonn which survived into NJ, e.g. saki-dat- 'ahead2

The first EMJ poetry anthology Kokinshii (914) stands out by having a single onbin fonn, tika-u ('close-ACOP.INF'), which is however used in a peculiar context, as an allusion to the flower kitikau.

7.1 Syllable structure

195

set.out; precede' and not the attested onbin fonn sai-dat-. A number of onbin fonns are attested only in kunten glosses or dictionaries, but were not used in general writing. Finally, in some cases both onbin and source fonns survived. This includes the establishment of grammatical alternation, such as between infinitive and onbin stem of the consonant base verbs, but also lexical items with some semantic or other specialization between the source and onbin fonn. A lexical example which illustrates the different possibilities is OJ kwo-miti 'small-road', which gave the attested onbin variants EMJ koudi and kondi. Of these, kondi did not survive into NJ, whereas both koudi > NJ koozi 'narrow backstreet' and kwo-miti > ko-miti 'small road' are used today, but with semantic specialization.

7.1.4.1 Onbin as sound changes Of the various forms which in Japanese philology are identified as onbin, it is possible to identify a set of core changes which can be said to constitute the 'onbin sound changes', i.e. a core of changes of certain CV syllables to bound moraic phonemes. This was preliminarily shown in Table 7.2 (a) above and is summarized here in (1 0). These changes account for the great majority of early examples of changes of lexical words. It was these changes which formed the basis for the accompanying change in phonological structure and whose outcomes served as a model for the morphophonological changes involved in onbin. Changes which are usually classified as onbin but are not included here are either sporadic changes which yielded fonns with one of the bound moraic phonemes, but generally are late and singular, listed in Table 7.2 under (c), or morphophonological developments such as those in (b) in Table 7.2 which are only found in the verb paradigms; such developments are analogical formations (see 8.1.4). It is traditional to distinguish four types of onbin: I-, U-, N-, and Q-onbin, depending on the bound moraic phoneme which resulted from the change, but we expand that to allow for the distinction between II, U/ and Ji, U/. As mentioned above, some source syllables gave variant outcomes, either a consonant or a vowel. (10)

Source syllable -pi, -pu > -bi, -mi, -mu, -gu > -ni, -nu > -bu> -ki > -ku> -gi >

Moraic vowel Oral Nasal

u

fJ fJ I

u

i

Moraic consonant Oral Nasal

Q N N

196

7 EMJ phonology

The source syllables thus consist of one of the consonants /p, k b, g m, n! and one of the high vowels !i, ul. As mentioned above, (7.1.4), syllables with OJ /Cwi/ (i.e., /pwi, kwi, bwi, gwi, mwi/) do not appear to have given onbin, with one or two exceptions. 3 In other words, it was only OJ /Cui and /Ci!, not /Cwi!, which gave onbin. A phoneme following the syllable that gave onbin was always a true consonant, /p, t, k, s, b, d, g, z, m, n/; that is to say, onbin did not take place before /r, w, y/.

7.1.4.2 Syllable reduction The onbin changes are sometimes described as, or illustrated by, segment loss and consonantal assimilation, so that for example takume is said to have become taume through simple loss of /k/; or pimuka(si) > piiiga through loss of /m/, and> pingasi through loss of lui, in which case the alternative outcomes are explained by the loss of different segments. It is possible to state the correspondence between source form and onbin form in this way in some cases, but not for example /pi/> lUI, or fbi, mi/ > 10!. A more comprehensive and realistic account describes these changes as phonemic reinterpretations of a phonetically reduced or weakened realization of a CV syllable as the realization of a single segment. This is illustrated in (11) which shows three stages of realization of the syllables involved. Column (a) shows the phonemic shape of the source syllable. (b-e) show phonetic variation which was regular in OJ: in both of them are shown intervocalic voicing of tenues and prenasalization of mediae (cf. 2.2.2), and tonality adjustment (2.3), and (c) alone shows continuousness variation among /p, k, b, g/ (2.2.3). In column (d) is shown a phonetically further reduced realization in which the CV syllable is realized as one unit with little contrast. The onbin sound changes thus consist in the reinterpretation of the realization in column (d) as deriving from a different phonemic representation than that in (b) and (c), shown in the final column, (e). The kind of phonetic reduction shown in (d) is typical of languages, like OJ, whose main or only syllable structure is CV; in particular, neutral or high vowels are usually reduced except in slow, distinct or elaborate diction, as is the case in NJ, especially as long as they leave a trace in the preceding consonant, for example in the form of palatalization or labialization. The more familiar or frequent in use a word is, the more likely it is to be phonetically reduced (this general principle is known as Zipf's Law). Many of the words which gave onbin were complex forms which at the same time were 3

An apparent exception is tuitaJi 'the first day of the month' which is usually said to be from *tukwi-tati 'moon-rise' (which is itself not attested in OJ). It is possible, however, that this is a folk etymology and that tuitati instead derives from a compound of tuki 'arrive'+ taJi 'depart' (notionally paralleling the name of the month January, named after the double-faced god Janus in Roman mythology). If so, *tukitati > tuitaJi is no exception.

7.1 Syllable structure

197

univerbated, e.g. kwo-miti 'small road' > koiidi 'narrow backstreet', and this fits the same pattern. Whether individual words gave onbin will have depended on their individual use, and this helps account for the fact that the onbin changes were non-automatic, yet pervasive. The role played by familiarity, and speech tempo and style, also makes it easy to understand the stylistic variation and distribution among unchanged and onbin forms of the same words. (11)

(a)

1-pul /-pi/ /-ki/

1-kul /-gi/

1-gul /-bi/

1-bul /-mil

1-mul /-ni/

(b) => => => => => => => => => => =>

[bwu [bji [gji [gwu (Ugji [UgWU [mbji [mbwu

-

-

-

[mii] [mwu] [nii]

(c)

(d)

pwu] .l3ji] Yii] ywu] uyii] uywu] mpji] mpwu]

ww:]

Wd [yd [yw:] [Uyd [UyW:] [m_l3d [m_[3w:] [md [mw:] [nd

(e) > > > > > > > > > > >

IU, Q/ IU, Ql Ill lUI

Iii !O,N/ !O,N/

!01 !O,N/ !O,NI IN/

7.1.4.3 Nasality As shown in (10) and (11), syllables with initial tenuis, /-pi, -pu, -ki, -kul, gave oral onbin, 11, U, Q/, whereas syllables with nasal or media, /-bi, bu, -gi, -gu, -mi, -mu, -ni, -nul, gave a nasal moraic phoneme, li, 0, N/, reflecting the phonetic nasality of the reduced realization which in the case of the mediae derived from prenasalization. As shown in the examples in Table 7.2, a tenuis in the source form changed to a media after a moraic phoneme deriving from /-bi, bu, -gi, -gu, -mi, -mu, -ni, -nul, e.g., pumite 'brush'> punde and pumite 'stepping' > punde. However, there was a difference in the phonemic shape of such onbin forms, depending on the morphemic constituency. As discussed above (7.1.2), nasality was not phonemic for moraic phonemes in morpheme internal position~ thus, in the course of the onbin sound changes, the phonetic nasality of the reduced realization of the source forms (11 d) was phonem icized differently depending on whether it was in morpheme final or morpheme nonfinal position, see (12), which shows both the phonemic shape and phonetic realization of (a) the source forms and (b) the resulting onbin forms. For the forms in (b), which have a morpheme boundary after the moraic phoneme, the phonetic nasality was assigned phonemically to the morpheme final moraic phoneme (/puN- + -tel) and the phonetic realization was derived as described in (8b) in 7.1.2.4. However, for the forms in (c), which have the moraic

198

7 EMJ phonology

phoneme in morpheme internal position, the phonetic nasality in the source form in (a) was in the course of the onbin sound change re-interpreted as deriving from a (prenasalized) media, but the moraic phoneme itself as being underspecified for nasality, i.e. /puCde/ whose phonetic realization was derived by the rules described in (8a) in 7.1.2.4.

(12) (i)

(ii)

a. {/pumi +tel 'step+ GER' {/tugi +tel 'follow+ GER'

b. => [pilmjAe]}

>

{[pfuldje] <= /puN + te/}

=> [tuuyjAe]}

>

{[tiii.ndje] <= /tui+te/}

{/pumite/ 'brush' => [pilmjAe]} {/tugite/ 'sequence'=> [tuuyjAe]}

> >

{[pfuldje] <= /puCde/} {[tiii. ndje] <= /tulde/}

Thus, phonemically the diachronic correspondences involved in onbin differ a great deal depending on whether the resulting moraic phoneme occurred in morpheme initial or non-initial position: /pumi +tel> /puN+ tel, but /pumite/ > /puCde/. However, in both of these cases we speak of /mi/ as having given N-onbin, in the one case (morpheme final position: /pumi + tel > /puN + tel) reflected as a nasal moraic phoneme, in the other (morpheme internal position: /pumite/ > /puCde/) reflected as a moraic phoneme underspecified for nasality but followed by a media which contributed the phonetic nasality to the surface form, through prenasalization and nasality spreading. And in both cases we use punde as a convenient transcription when we are not specifically concerned with the phonemic shape of the form (cf. 7.1.3). An important point about (12) is that it shows that the onbin sound changes involve the change of one set of phonemic representation and its phonetic representation, shown in column (a), into another set of phonemic representation and its phonetic realization, column (b). Thus, stating the change as {/pumite/ 'brush'=> [pilmjAe]} > { [pfindje] <= /puCde/} says a great deal more than does a simple statement of the phonemic diachronic correspondence, such as /pumite/ > /puCde/. Some examples in OJ of syllable loss seem to involve the same kind of phonetic reduction as was involved in onbin, see (13), cf. 2.7.1.1. The main difference between the developments in (13) and in (12) is that no moraic phoneme arose in the examples from OJ in (13). This suggests that the phonetics which in the transition between OJ and EMJ gave rise to onbin already were a feature of OJ. Similar reductions took place through EMJ, e.g. the copula gerundnite >de (8.2) or nani-to > nado (8.7.3). The morphophonological process of rendaku is also thought to originate in similar reduction of particles such as genitive no and dative ni (2.6.2).

7.2 The soWld shape of SJ vocabulary

(13)

{/*ami-piki/ 'net-pulling' => [ami:biig}]} > {/kigisi/ 'pheasant' => [klY/3i]} >

199

{[amb}g}] <= /abiki/} {[kr3i1<= lkizil}

7.1.4.4 Major class; consonant or vowel? An interesting aspect of the onbin changes is that some source syllables gave both vocalic and consonantal outcomes. The basis for this is that the reduced realization in (11d) in some cases lent itself to interpretation as both a vowel and a consonant. It is not possible based on the written sources to determine any systematic distribution of the consonantal and vocalic outcomes. The consonantal outcome is often assumed to have been characteristic of eastern dialects. This is quite possible, and fits in well with an overall characterization of western Japanese as more vocalic and eastern Japanese as more consonantal, as reflected today for example in western NJ having less vowel devoicing than eastern NJ; however, for onbin this characterization is to some extent based on the present-day distribution of verbal onbin forms, which up until the end of the LMJ period was not as clear-cut as it is today (see 12.3.2). 7.1.4. 5 Other sources of bound moraic phonemes Apart from onbin, II, U/ result from the regular loss of /p/ and /w/ before Iii and /u/ which took place between the end of the tenth century and 1100, see below (7.3.1.1, 7.3.2.3). As mentioned below, bound moraic phonemes, both consonants and vowels, were frequent in SJ vocabulary (7.2). Both the onbin sound changes and the introduction of syllable length are often ascribed to influence from SJ. This is, however, doubtful, because of the fact, amongst others, that onbin forms and SJ vocabulary occupied different ends of the stylistic spectrum. It is rather the case that incorporation of SJ loanwords was facilitated by the emergence of the new syllable structure and phonemes. Thus, early loans from Chinese in OJ reflect EMC *-IJ by /gV/ (e.g. EMC *~aiWIJI ~a:iWIJ (l() >OJ sugu 'double'), but later by !0, I/ (cf. the SJ readings of l(: sou, sail; see 9.2.2.3). 7.2

The sound shape of Sino-Japanese vocabulary

The intake of SJ loanwords resulted in the introduction of some sound combinations and distributions which were initially only used in the SJ vocabulary. First of all, there were no restrictions on the occurrence of word initial mediae or liquid, that is to say, SJ loanwords freely had initial b-, d-, g-, z-, r- (see e.g. examples in 9.2.3.2 below). Second, as shown in Table 7.1 above, SJ loanwords (re)introduced complex syllable onsets, ICy-! or /Cw-/: /Cya, Cyo, Cyu; Cwa, Cwe/. Note that on the phonemic interpretation of the kO-otsu distinctions adopted in this book, also OJ had complex

200

7 EMJ phonology

syllable onsets, but with different combinations of glides and vowels: /Cye, Cwi, Cwo/ ( cf. 2.1.4). Of the new complex syllable onsets introduced by SJ loanwords, /Oja-1 was the most frequent. With the exception of a num her of words with /syu, syo, zyu, zyo/, other /Cyu, Cyo/ were rare in EMJ, although they are frequent in NJ; they are included in Table 7.1 in brackets because they are found in a few SJ words taken in towards the very end of the EMJ period. SJ /Cyo/ occurs in /Cyoku/ and a few words with /Cyofil (:: EMC *-iJJ) in words mostly taken in during LMJ. Most other instances of NJ SJ /Cyu, Cyo/ are found in the outcome of LMJ contractions of EMJ /Cyaii, Ceu, Ciu, Cipu/ (see 11.5). Through sound changes in LMJ, /Cya, Cyo, Cyu/ also came to occur in the native vocabulary. /Cw-/ is only found in /kwa, gwa, kwe, gwe/; /kwe/ is found in a single native word, the verb kwe- 'kick' (see 8.1.2). In general, SJ loanwords are characterized by a high frequency of long syllables. In addition to the bound moraic phonemes II, U, i, 0, N, Q/, some SJ words had syllable final 1-t/ (reflecting EMC *-t, see 9.2.2.3). There was no ready way in the kana script of noting syllable final /-t/ and the main evidence is from forms exhibiting renjo (see 7.2.1 immediately below) and in late LMJ transcriptions, such asfotnet suru 'get fever'. Syllable final 1-t/ will be further discussed in 11.4. Furthermore, SJ words are often said to have had a distinction between syllable final /-n/ and /-m/ (reflecting EMC *-nand *-m). The sole evidence invoked for positing syllable final /-m/ is from forms exhibiting renjo, but it seems likely that the distinction between 1-ml and 1-n/ was in fact never a part of Japanese, see below (7.2.1).

7.2.1

Renj6

Renjo (J!F' 'liaison') is a phonological process of onset creation which originates in SJ. When a syllable final consonant was followed by a syllable with initial vowel or glide, the consonant was carried over to create the onset in the next syllable. An early example from (some versions of) the Genji monogatari is onmyauzi 'fortuneteller' (> cNJ onmyoji) which reflects Chinese llii~~ffi whose individual SJ kanji readings are on, yai1 (> cNJ yo), and si. The form onmyauzi thus both shows creation of an onset in the second syllable, as well as suggesting that the reading of the first kanji ended in -m, as in EMC: *?im. Sanmi 'third rank' (san 'three' ~ :: EMC *sam + wi 'rank' {ll.) is a similar example. However, rather than positing distinct syllable final /-m/ within the SJ sound system on the basis of a few examples like these, it seems more likely that forms such as onmyauzi or sanmi were taken in as whole words from the Chinese used in Japan(' Japano-Chinese') and that there never was a distinction between 1-ml and /-n/ in SJ. See 9.2 about the distinction between Japano-Chinese and Sino-Japanese.

7.3 Regular segmental soWld changes

201

Renjo is usually thought to have been a regular, automatic feature of pronunciation which was therefore not noted much in writing. It ceased to be productive early in NJ where it only survives in a small number of lexicalized forms, see (14) which gives the cNJ form and the EMJ form if the two are different, as well as the cNJ and EMJ readings of the individual kanji. Note that /C.wV/ => /C.CVI, with loss of the labial glide. (14)

tenno 'emperor'< tennaii(~ ten+~ 8 <wail) ninnaji < ninnazi 'Ninnaji' C1= nin + fiJ wa + ~ zi) rinne 'transmigration (of souls)'(~ rin + ~ e <we) hanna < pannoii 'reaction' (R han <pan + ~ 8 < oil) kannon < kwannon 'Kannon' (iii. kan < kwan + if on) innen 'destiny' ([Z,;I in+~ en) annon 'peace and quiet' (:i( an+ ~ion) ginnan 'gingko nut'(~ gin+~ an) kanno < kannoii 'sympathetic response'(~ kan + ~ 8
7.3

~

in)

Regular segmental sound changes

Two overall sets of sound changes took place during EMJ: the merger of intervocalic /p/ with /w/ (7.3.1.), and the loss of (phonemic) glides before /e, i, o/, including the loss of the ko-otsu distinction (7.3.2). These changes are all mergers and are textually manifested as loss of orthographic distinctions. An overall summary of sound changes which took place during this and following periods is given in the appendix.

7. 3.1

Changes affecting OJ /pi

One of the main issues in Japanese historical phonology is the complex of changes which have affected OJ /p/, from EMJ to NJ. They are summarized in (15): (15) a. 1-p-1 > 1-w-1 > 0 before all vowels except Ia/. b. /p/ >If/ > /h/(realizedas[p]before/u/andas[<;]beforeli/) The first change, of intervocalic 1-p-1 to 1-w-1 took part in EMJ and will be discussed in this chapter. After that merger 1-w-1 was lost before li, e, o/, see 7.3.2.3. The second change, of /p/ to If/, appears to have taken place in LMJ

202

7 EMJ phonology

and will be addressed in 11.3, as the first positive evidence we have for the stage lfl is in the Christian materials from the end of LMJ, but we will in this chapter briefly review evidence for the sound value lpl in EMJ of the reflex of OJ lpl (7.3.1.3). The change of lfl >/hi took place during NJ, see 14.3. 7.3.1.1 Merger of intervocalic /-p-I with 1-w-1 Some time in the second half of the tenth century, intervocalic 1-p-1 changed to 1-w-1 (merging with earlier 1-w-1), except before lui where 1-p-1 was lost, see (16). This change resulted in the merger of previously distinct forms, exemplified in (17). Although this sound change properly dates from the second half of the tenth century, early word-individual examples are seen already at the beginning of EMJ. A well-known early example is the adjective OJ uropasi 'splendid' which already in some early EMJ materials is spelled urowasi.

(16)

(17)

p

>

wl 01

V_ {i, e, a, o} u

v

OJ apa 'millet' f. awa 'foam' api 'meeting' f. awi 'indigo' kapo 'face' f. awo 'blue' upe 'top' f. uwe 'planting' apu 'meets'

> > > > >

EMJ awa awi (k)awo uwe au

OJ did not have the syllable lwul. The fact that 1-p-1 in the course of this sound change was lost before lui, and did not give /wul, indicates that the absence of the syllable lwul was no accident, but the result of an active phonotactic restriction. The lui which resulted from 1-pul was incorporated into the preceding syllable as the second mora of a long syllable. 1-pul > 1-ul was frequent in SJ vocabulary, where early EMJ 1-pul reflected EMC syllable final *-p (see 9.2.2.3), e.g. early EMJ sipu (#:) 'anthology' > late EMJ siu, tapu (~) 'stupa' >tau. The sound change 1-p-1 > 1-w-1, merging intervocalic 1-p-1 with the already existing phoneme 1-w-1, is a straightforward sound change, which is observable and dateable through the written sources as orthographic confusion of kana for pV and wV syllables in medial position. Its background is the phonetic realization of tenues (/p, t, k, sf) mentioned in 2.2.3, so that intervocalic 1-p-1 already in OJ was phonetically non-distinctively voiced and had both occlusive and fricative variants: 1-p-1 => [b- .(3]. The phonemic change consisted in the reinterpretation and phonemic identification of these phonetic realizations of 1-p-1 with 1-w-1, due to the phonetic similarity of [13] and [w ], likely related to

203

7.3 Regular segmental soWld changes

the fricative realization of 1-p-1 gradually becoming more prominent or frequent. (18)

1-p-1 => 1-w-1 =>

[b- 13] [w]

}>

1-w-1

/apal => lawai=>

[aha- a_l3a] [awa]

}>

lawai

This scenario makes it easy to understand the appearance of early wordindividual instances of this change, such as urupasi > uruwasi: as long as 1-p-1 and 1-w-1 coexisted, individual words could be reinterpreted as having 1-w-1 rather than 1-p-1, with no effect on the phonological system. Although all intervocalic tenues were phonetically voiced, and it is likely that /k/, like /p/, had fricative variants, /k/ or It/ did not undergo similar changes, as there was no scope for reinterpreting their phonetic realizations as realizations of other already existing phonemes; thus, [-g-- -y-] and [-d-] remained uniquely identifiable as realizations of /-k-/ and 1-t-1. As the mediae, /b, d, g, zJ, were prenasalized, [mb, nd, Dg, nz] (cf. 2.2.2), they were readily distinguishable from the intervocalic phonetically voiced realizations of 1-p-, -t-, -k-, -s-1. It was earlier thought that the change 1-p-1 > 1-w-1 proceeded through the EMJ period as something like (1-p-/ [p] >) 1-p-1 [p] > 1-w-1. On that view, the change mainly consisted in the introduction of intervocalic voicing of 1-p-1 to 1-w-1, but not of other tenues. However, it has, since Giinther Wenck proposed this in the early 1950s, become increasingly accepted that OJ tenues were phonetically voiced intervocalically, and following on from that that 1-p-1 > 1-w-1 did not have the intermediary stage 1-p-1 [p], but that the change proceeded as in (18).

7.3.1.2 Retention of 1-p-1 1-p-1 is retained after the moraic oral obstruent /Q/. A few emphatic words have preserved 1-p-1 after /Q/ into NJ, e.g. appare 'splendid, brilliant; bravo!' (OJ apare 'wow!; being evocative of deep feeling' which also has the reflex aware, from Heian times a central element in Japanese aesthetics), moppara 'solely, entirely' (OJ mopara). It is likely that these words existed as emphatic forms with a forcefully articulated /p/ already in OJ and were phonemicized with /Qp/ after the emergence of /Q/. Among SJ vocabulary the retention of /p/ after /Q/ has been regularized so that all NJ SJ morphs with initial /h-/ ( < /p/) have automatic allomorphs in /-p-/ for use after /Q/. Similar alternations are preserved in the native vocabulary, so that for example NJ verbs in initial /h-/ have /p-1 when combining with the prefix hiQ- (< EMJ piC-
204

7 EMJ phonology

lexicalized and some may have been fonned analogically, but the point remains that they show preservation of /p-/ after /Q/. When /p/ was initial in the second component of a transparent compound it did not change to /w/, but remained /p/. All such remaining cases of /p/ later became LMJ /f/ > NJ /hi (see further 11.3, 14.3; for example OJ asa-pi 'morning-sun'> LMJ asa-fi > NJ asa-hi. Proper names present an interesting case. Some, such as early EMJ Pudi(-)para > late EMJ Pudiwara > NJ Fujiwara, have 1-p-1 > 1-w-1 and therefore seem to have been univerbated, whereas others such as EMJ Naka-para > NJ Nakahara, or Kiyo-para > Kiyohara behave like compounds and have 1-p-1 > If/> NJ /hi, although all involve a reflex of OJ para 'plain, field' as their second element. The presentday name of the supposed main character of the Ise monogatari, Ariwara no Narihira (825-80), which in his day would have been Aripara no Naripira, exhibits both types of development. Apart from a negligible number of exceptions, such as bipa (~~) 'flute'> biwa, SJ vocabulary with a second component in initial /p/ were treated as transparent compounds in which 1-p-1 did not change to 1-w-1, but was left as /p/ which eventually gave LMJ /f/ > NJ /hi, for example OJ/EMJ mupon (~~, EMC *muw puan'; only logographic OJ attestation) 'rebellion, treason'> LMJ mufon > NJ muhon.

7. 3.1. 3 The Early Middle Japanese sound value of the reflexes of initial /p-1 Until fairly recently the widely accepted consensus view was that pre-OJ /*p/ already by, or in, OJ had changed to become /pl. There is, however, little to support this view. With some regular exceptions (see 11.3), the reflex in NJ of pre-OJ initial /*p-1 is /h/, realized as [p] before lui and as [c;] before Iii. The Christian materials from the end of the LMJ period where the reflex of OJ /p-1 is transcribed 'f', e.g.fito 'person' (OJ pito, NJ hito),fana 'nose' or 'flower' (OJ pana, NJ hana), show that an intennediary stage /f/ (possibly realized as [!}>])existed between pre-OJ /*p/ and NJ /h/: /*p/ >If/> /hi. Apart from the terminus ante quem provided by the Christian materials, the change of /p-/ > If-/ is very difficult to date for it resulted in no mergers and found no orthographic expression. The only materials available are evidence from the sound values of man 'yogana, internal alternations, some notes on pronunciation from the EMJ period, considerations of the kana orthography, and foreign transcriptions, primarily the Christian materials. First of all, it is no longer controversial to regard the reflex of pre-OJ /*p/ as having been /p/ in OJ. That is amply shown by the reconstructed EMC sound values for the ongana used for those syllables, as well as by phonological alternations between OJ /p/ and /b/. The earliest indication which is taken

7.3 Regular segmental soWld changes

205

as positive evidence for a sound value [cp] or [f] for the reflex of OJ /p-1 are articulatory descriptions or instructions in the priest Shinren's Shittan kuden and Shittan soden, both from the very end of the Heian period. This material has traditionally been taken to show that the reflex of OJ /p-1 at that time was pronounced as a bilabial fricative or approximant, but the fact is that Shinren's descriptions are obscure and difficult to interpret 4 Furthermore, as Kiyose (1985) points out, there are strong internal arguments in favour of viewing the EMJ reflex of OJ initial /p-/ as having been /p/ throughout that period and probably well into the LMJ period. The main facts are (a) that the reflex of OJ /p-1 was paired with /b/ in exactly the same way It, k/ were paired with /d, g/ in sound changes and phonological rules through the EMJ period~ and (b) that single kana letters were used to represent both /p VI and lb VI syllables, for example l"i used to represent both /pa/ and /ba/, in exactly the same way that iJ., represented both /ka/ and /ga/ and t-c. both Ita/ and /da/ (6.1.2.1)~ this shows that /p/ and /b/ were phonologically paired as tenuis and media at the time the letter categories of the kana writing became fixed. Finally, when the change /p/ > /f/ took place, /p/ was preserved after IN, Ql, providing further support for /p/ having remained /p/ into the LMJ period, see 11.3.

7. 3.1.4 Summary Summarizing, OJ /p/ phonemically remained /-p-/ in intervocalic position (with voiced and both occlusive and fricative variants [b - j3]) in EMJ until its merger with 1-w-1 in the tenth century. Furthermore, we follow Kiyose that initial /p-/ phonemically remained /p/ (with both occlusive and fricative variants [p- cp]) through the entire EMJ period and (well) into the LMJ period. Sometime before the end of the LMJ period /p/ > /f/ (either a bilabial fricative [cp] or a labio-dental [f])~ this change will be discussed in 11.3. 7. 3.2

Loss of labial and palatal glides

In the course of a set of changes which took place between late OJ and around 1300, the palatal glide /y/ was lost before /e/ and the labial glide /w/ was lost before /o, i, e/.

4

The relevant passages are (cited from Kobayashi 1981): Shittan kuden (~\1;0~): in the articulation of"pa", use -'lr*l?t 'the inner part of the lips', but in the articulation of"ma", use !l5'f.:7.)- 'the outer part of the lips'. Shittan soden (~\l;tll~): to articulate "pa': l:T-'l~~ !llfailii!Vtff 'Join the upper and lower lips and say out "a" softly/gently. Then this sound will result'; and for "rna": M-'l:91--l~Mi!llfa(ilii)!Vt{;f 'Close the outside of the lips and say "a" very forcefully. (Then) this sound will result.' Much speculation has gone into interpreting these instructions and they are conventionally taken to mean that "p" was a bilabial fricative [cjl], but this is far from clear.

206

7 EMJ phonology

7.3.2.1 Merger ofko-rui and otsu-rui syllables; loss of post consonantal ,YI before lei and lwl before li, ol The first of these changes is a loss of the distinction between the ko- and otsurui syllables (cf. 2.1). These mergers set in towards the end of OJ and are in the main thought to have taken place in the transition between OJ and EMJ, but the distinction between Cwo and Co was retained longer in some contexts, notably for kwo-::/: ko into the tenth century (in texts written in man 'yogana). On the phonemic interpretation introduced in 2.1.4 above, and reflected in the transcription used in this book, these changes may simply be understood as the loss of a glide between the onset consonant and nuclear vowel in complex syllables, see (19). The consonant in /Ce, Ci/ which resulted from these mergers was palatalized ([qe, Cii]); palatalization before /e, if was a feature already of OJ (cf. 2.3) and it was maintained into the NJ period before /e/ and is still found today before Iii (cf. 2.3; 11.6.1, 14.4.1). (19)

me 'eye'-::/: mye 'woman' pi 'sun'-::/: pwi 'fire' ko 'this'-::/: kwo 'child'

> > >

me pi ko

7.3.2.2 Loss of syllable initial ,YI before lei Next, syllable initial /y/ was lost before the front vowel /e/ during the first half of the tenth century, e.g. (20). This change is thought to have been complete by 950 after which man'yogana and kana for /.ye/ and /.e/ were no longer kept distinct. Recall that the Ame-tsuchi no kotoba reflects a language stage when/. ye/ and/. e/ were distinct, whereas the language reflected in the Taini-uta and the Iroha-uta no longer had that distinction (6.1.2.4). This change is usually thought of as a merger of the syllables /.ye/ and I.e/, but that only holds in word initial position, as OJ had no word internal N.e/ sequences (cf. 2.5). Note, importantly, that/.e/ resulting from the loss of phonemic /.y/ before /e/ was pronounced with an automatic phonetic onglide, as shown in (20); see further 7.3.2.4-5 below. (20)

e- 'to get' -::/: ye 'inlet' nuye 'mountain thrush'

> >

e ([ie]) nue ([nuie])

7. 3.2. 3 Loss of syllable initial lwl before lo, i, el Finally, syllable initial /w/ was phonemically lost before /o, i, e/. This happened in the following stages over several centuries and was not concluded until the LMJ period. These changes took place after intervocalic 1-p-1 > 1-w-1 (7.3.1.1) and they therefore also affected 1-w-1 < 1-p-1. As opposed to the loss of post-consonantal /y, w/ and of syllable initial /y/ before /e/, the loss of syllable initial /w/ took place after the establishment of the letter categories of the Iroha and it is therefore ignored in the etymological kana spelling which retains several separate letters for each of /.o, .i, .e/ (6.1.3).

207

7.3 Regular segmental soWld changes

Syllable initial /w/ was lost before /of, completed around 1000. This change is usually thought of as a merger between the syllables /.wo/ and /.o/, but as with the loss of /.y/ before /e/, that only holds in word initial position as OJ had no /Vo/. Also the outcome of this change was pronounced with an automatic phonetic onglide (see 7.3.2.4-5). (21)

oto 'sound' -=F wotoko 'man' awo 'blue' (kapo >) kawo 'face'

> > >

oto(ko) ([w odogo]) ao ([awo]) kao ([kaw o])

The loss of /w/ before li, e/ proceeded in two phases: in word medial position, concluded around 1100, e.g. (22); in word initial position, concluded around 1300, i.e. in the LMJ period, e.g. (23). (22)

mawirn 'comes, goes (HUM)'> (mapi >) mawi 'dance'> (OJ opwi > EMJ opi >) owi 'grow.INF' > uwe 'plant.INF' > (upe >) uwe 'top'> (OJ mapye > EMJ mape >) mawe 'front'>

mairn mai oi ue ue mae

(23)

winaka 'countryside' > wemi 'smile' >

inaka emi

7.3.2.4 /-i, -ulversus I-.e, -.o/ The changes outlined above resulted in a large number of instances of vowel sequences (/-Vi, -Vu, -Ve, -Vol) from different sources, see (24). Recall that vowel sequences were not allowed by OJ phonotactics (cf. 2.5).

(24) -Vi

<

-Vu

<

-Ve

<

-Vo

<

-Vwi -Vwi -Vwi

< <

-Vpi -Vpi

<

-Vpwi

-Vye -Vwe -Vwe -Vwe

< <

-Vpe -Vpe

<

-Vpye

-Vwo -Vwo

<

-Vpo

-Vpu

208

7 EMJ phonology

In EMJ, li, uJ were incorporated into the preceding syllable to fonn the second mora of a long syllable, just like the II, U/ resulting from the onbin changes (7.1.4) and were later subject to various contractions (11.5). On the other hand, /e, o/ maintained their independence as syllable initials and were pronounced with an automatic onglide which created a phonetic onset to the syllable: [ie] and [wo], respectively; this was also the case in word initial position, (25). (25)

e- 'to get' eda 'branch' emi 'smile' nue 'thrush' koe 'voice' ue 'top'

[ie] [ienda] [iemii] [nuie] [koie] [uie]

(
oto 'sound' otoko 'man' ao 'blue' kao 'face'

[wodo] [wodogo] [awo] [kawo]

(
This feature of pronunciation is still clearly reflected in the Jesuit texts from the end of the sixteenth century (11. 8) and was maintained into the NJ period (14.4), for /.e/ late enough to be reflected in for example the English name, yen, for the Japanese currency (NJ en). The palatal onglide preceding I.e/ was phonetically, if not functionally, related to the palatalization of consonants before /e, i/ (cf. 7.3.2.1 above). 7.3.2.5 Phonemicization: I-.e, -.o/ or 1-.ye, -. wo/ The question arises why the glide in fonns such as [taie] and [awo] is interpreted as being automatically inserted and not as being present in phonemic representation. This is a particularly pertinent question because there were no words in OJ of the structure /CVe/ and /CVo/, so in word internal position no merger took place between /.ye/ and /.e/, or between /.wo/ and /.o/. In other words, why do we interpret these fonns as /tae/ (=> [taie]) and lao/(=> [awo ]), and not simply as /taye/ and /awol, which is the shape they had in OJ? This question can be viewed as a theoretical one which will receive different answers in different phonological frameworks, but from a more practical point of view several factors suggest that word internal, syllable initial /y/ was lost before /e/, and /w/ before /o/, at the same time as in word initial position, i.e. (26), even if these changes in non-initial position resulted in no mergers and are not manifested as a loss of orthographic distinctions.

(26)

/Vye/ > /Vwo/>

/Ve/ !Vol

7.3 Regular segmental soWld changes

209

First of all, the sound changes that were discussed above form a clear pattern, with loss of /y, w/ before certain vowels. The loss of post-consonantal /y/ before lei and /w/ before /o/ (7.3.2.1), and of word initial /y/ before lei and /w/ before /of (7.3.2.2-3), is then paralleled by the loss of word internal, syllable initial /y/ before lei, and /w/ before /o/. They may all be captured simply as (27), with no reference to position in the word or syllable: /y/ > 0 I _ lei /w/ > 0 I /o/

(27)

Second, /w/ has gradually been lost before other vowels than /a/ (11.2) and /w/ is thus only preserved in the environment of maximal contrast; it would be difficult to explain the retention of phonemic /w/ before /o/ later than before li, e/ where the contrast is greater than before /o/. Conversely, the phonetic onglide is functionally well motivated, providing a phonetic syllable onset to a vowel initial syllable. Third, as described above, also /w/ changed before lei, merging [taie] 'cease' lye/, which while not impossible is less straightforward than either of lye/> lei and /we/> lei. Fourth, observe the verb forms in (28) (see further 8.1.2). The infinitives are identical (in their last syllable), whereas the conclusive of (a) is different from (b) and (c). As shown, verbs in (a) reflect OJ stems in C~e whose regular conclusive is OJ and EMJ C~. This shows that the (a) verbs have underlying stems in //CVye// whose palatal glide surfaced when followed by lui. On the other hand, the stems in (b) reflect OJ CVpe and those in (c) reflect OJ CVwe, which changed in the course of the sound changes above, merging the shape of the infinitive with the reflex of OJ C~e, as part of the overall changes which merged /Vye, Vwe, Vpe/. However, the conclusives are different, with the (b) and (c) verbs showing no trace of a glide. That suggests that they underlyingly had no glide, as opposed to the verbs in (a), and that the glide in the infinitive is the result of automatic glide generation. (Much later, in late LMJ, all three verb base types merged, see 12.3.1.) (28) a. b. c.

'cease 'exchange' 'starve'

Infinitive [taie] [kaie] [uie]

Conclusive tayu kau uu

EMJ base < OJ base //taye// taye//kae// kape1/ue/1 uwe-

Finally, in LMJ /au/ was contracted to a long vowel [~:]. When /.au/ was syllable initial, the outcome of the contraction was [w:>:], as shown by the

210

7 EMJ phonology

transcription in the Jesuit texts where the contracted form for example of au 'meets' is written vo (see 11.5). In this case the glide is clearly automatically generated.

7.4

Prosody; 'accent'

NJ is usually characterized as a 'pitch accent' language. Certainly, the great majority of Japanese dialects employ pitch phonologically, and Tokyo (Standard) NJ has a prototypical pitch accent system, with distinctive lexically specified accent, which is realized by pitch, as well as accentuation rules assigning accent to inflected and other complex forms. This is also the case for modem Kyoto, whose system, however, is more complicated, involving both accent and tone, more narrowly defined. And for some dialects, for example Kagoshima, the phonological system of pitch seems to be less, if at all, accentual than tonal. In Japanese, the phonetics and phonology of pitch is customarily referred to as akusento, and this usage has to a large extent carried over into English language scholarship as 'accent', but in order to avoid potential misunderstandings, we refer to the phonology of pitch as 'prosody', distinguishing 'accent' and 'tone' as appropriate. Some dialects, usually referred to as 'accentless', do not employ pitch distinctively and generally words in those dialects have flat pitch contours. Prosody plays a big part in Japanese dialectology. Three main types of systems using pitch distinctively have been identified, with Tokyo, Kyoto and Kagoshima being typical representatives, but with much variation within the systems. The basic descriptive facts of the three types of systems are as follows. Kyoto distinguishes four basic word melodies, [high], [falling], [rising], and [rising-falling]. Phonologically this manifests combinations of lexical wordtone (defined by the initial pitch of a form) and accent (defined by a drop in pitch). Words with [high] or [falling] melodies, which begin on a high pitch, have a lexical High wordtone (H-words), those with [rising] or [rising-falling] melodies, which begin on a low pitch, have a lexical Low wordtone (L-words). In Samuel E. Martin's terminology, wordtone is referred to as 'register'; see further about lexical wordtone 7.4.3. Words with a drop in pitch, [falling] and [rising-falling], have an accent on the last mora before the drop. The location of an accent is distinctive, but initial accent is allowed only in H-words and final accent only in L-words of two moras (or three-mora words of the shape /CVQ.CVI, e.g. makka LLH(L) 'deep red'). Lexical words of the shape HHL have recently changed to HLL, so in words of less than four moras the location of an accent is no longer distinctive (but cf. nadesiko 'pink (name of flower)' LHLL versus irogami 'coloured paper' LLHL). The accent bearing unit is the mora, but of the bound moras, /Q/ cannot be accented. Unaccented L-words have an automatic final rise in pitch. Noting High wordtone by a preposed

211

7.4 Prosody; 'accent'

acute, Low wordtone by a grave, and accent by an acute over the accented mora, we can note these forms as in (29), using "o" for mora: (29) Unaccented

High Accented

HH 00

arne 'candy' HHH 000

katati 'shape'

Low Unaccented

Accented

HL '6o hana 'flower'

LH

LF o6 arne 'rain'

HLL '6oo awabi 'abalone'

LLH

00

kasa 'urn brella' 000

usagi 'rabbit'

LHL o6o tokage 'lizard'

Tokyo distinguishes three basic melodies, [rising], [falling], and [risingfalling], manifesting unaccented versus initial or non-initial accented; thus wordtone is not relevant in Tokyo. The location of an accent is distinctive. The accent-bearing unit is the syllable; long accented syllables have a drop in pitch from the first to the second mora. Lexical final accent is allowed, but only realized when followed by a neutral particle, as shown in (30). Unless the first syllable is accented, the first mora in a word is phonetically [low]. (30)

Unaccented

Accented Initial

LH(-H) 00

arne 'candy' LHH(-H) 000

katati 'shape'

HL(-L) 6o arne 'rain' HLL(-L) 6oo awabi 'abalone'

Non-initial LH(-L) o6 hana 'flower' LHL(-L) o6o tarnago 'egg'

LHH(-L) oo6 atarna 'head'

Kagoshima distinguishes two basic word melodies [(rising-)falling] and [rising], often classified as 'A' and 'B', respectively. The location of a pitch drop or rise is not distinctive, but is always between the penultimate and final mora, moving if a neutral particle is attached. This system is often thought of in terms of wordtone, such that the [falling] melody manifests High or Falling tone and the [rising] melody a Low or Rising tone, but it may be analysed in different ways, for example in terms of accented versus unaccented, or as penultimate versus final accent. Here the forms are noted with preposed circumflex for High-Falling tone and grave for Low(-Rising) tone.

212

7 EMJ phonology

(31)

Falling (A)

Rising (B)

HL-LH-L

LH -LL-H

1\00

' 00

arne 'candy'

arne 'rain'

LHL-LLH-L

LLH -LLL-H

/\000

' 000

katati 'shape'

awabi 'abalone'

Despite the waves of standardization and dialect eradication and erosion which have taken place since the beginning of the twentieth century, dialectal prosodic systems and differences between dialects are well preserved, and prosody figures prominently in meta-linguistic consciousness about speakers and varieties of Japanese. Dialect groupings based on prosody to some extent yield different divisions than isoglosses based on segmental phonology, vocabulary, or grammatical features. Prosody has for that reason, amongst others, attracted great interest in reconstruction of proto-Japanese. Map 7.1 shows the distribution of the main types of prosodic systems. Much scholarship on Japanese prosody is couched in tenns of phonetic pitch patterns or melodies. Correspondences between dialects or between different language stages are often stated as relations between phonetic pitch patterns, e.g. HL :: LH, or LL > HL, with no regard to underlying phonological shapes. This facilitates comparison between structurally different systems, but sometimes obscures the differences between them.

7. 4.1

Prosodic classes

Based on correspondences between Japanese dialects, combined with the written sources noting pitch, Kindaichi Haruhiko (1913-2004), one of the pioneers of diachronic and historical research into Japanese prosody, proposed a reconstruction of the prosodic classes of pJ. The classes, which fonn the basic framework for work on Japanese prosody, are noted in the fonn 'x.y', where 'x' represents the number of moras and 'y' the class, such that for example '2.3' is class three among two mora nouns. The classes are shown below in Table 7.3 and representative members of each are given in (35). The reconstruction overall works well for the mainland dialects of Japanese. Different classes have merged differently in different dialects, but the classes exhibit fairly good correspondences both across time and between dialects, so that it is largely possible to predict the prosodic shape of a word in a dialect from its prosodic class (with the exception of inevitable word individual changes in class membership over time); for example class 2.3 nouns, e.g.

7.4 Prosody; 'accent'

213

Accent ua I ty pes ([[[[] Kyot o- Osaka ~ West ern Kyushu ~ Tokyo Accentless ~ One-pattern

c=J

. '

100

200 100

300 200

400

500 300

600 km

400 miles

Map 7.1 Distribution of prosodic systems (from Shibatani 1990: 211)

mimi 'ear' or yama 'mountain', have the pitch shapes HL in Kyoto, LH(-L) in Tokyo, and LH in Kagoshima. While the classes provide a reconstruction of the distinct pJ classes, they are largely established on the basis of correspondences between prosodic classes of living dialects and the words that populate them include both loanwords or transparent compounds, both of which of course do not belong to pJ. It should also be mentioned that classifications of Japanese dialect do not always make clear whether their criteria for classification are the phonological systems of the dialects, or the patterns of merger of prosodic classes exhibited by the dialects. Prosody thus plays a central part in the reconstruction of pJ and the charting of how different dialects have branched off, and in general in an overall

214

7 EMJ phonology

understanding of the pre-history of Japanese. Here, however, we focus on the attested changes in prosody, between the earliest sources and the present day, that is, modern Kyoto (or Kansai) Japanese which is the descendant of the language reflected in the OJ and MJ sources. Reference will only be made in passing to reconstructions of pJ or to Tokyo Japanese, whose contemporary prosodic system is well described, but of which no records exist before the NJ period. The first extensive sources for prosody are from the second half of the EMJ period and consist in materials which note the pitch of words by means of tone dots. The main source with tone dots is the dictionary Ruiju-myogi-shO, whose original date is 1081. Tone dots originate in China where they were used to mark the tone of a syllable by their position next to a kanji. This practice was first adopted in kunten materials to indicate the tone of Chinese words, but later adapted to note the pitch of Japanese words. As mentioned above (6.1.2.2), tone dots are in some materials extended to note the distinction between tenues and mediae, with pitch indicated by the position, but tenuis or media by the shape of the marks. The tone dots used in Japanese materials are as shown in (32), with four positions: lJLF' hyoshO 'even tone', Jl:tF' tosei 'east tone', 1:: F' joshO 'rising tone', and $:F' kyoshO 'departing tone', respectively. (32)

1::•

•-*

•· SJZ• The lJL and 1:: dots are overwhelmingly frequent. The $: dot represents a combination of lJL 1::, whereas the Jl:t dot is taken to represent the opposite sequence: 1:: lJL. The dots are thought to represent the following pitches: 5

' An ingenious alternative, chiefly developed and articulated by S. Robert Ramsey, reverses the reconstruction of the phonetic pitches of the !J!. and J:. dots, so that !J!. represents H, and J:. L (see Ramsey 1979a, 1979b, and more recently de Boer forthcoming). This proposal entails that EMJ in terms of phonetic pitches was similar to present-day Tokyo and that present-day Kyoto represents a dramatic innovation. Ramsey's ideas have, however, been rejected by most other scholars.

7.4 Prosody; 'accent'

(33)

215

1::. [H]igh pitch

lJL [L]ow ll: [F]alling (= [HL]) :;iS [R]ising (= [LH]) Earlier materials have been shown to have differentiated otherwise segmentally equivalent phonograms (man 'yogana or kana) to distinguish between high and low pitched syllables or moras, for example in some parts of the Nihon shoki. This usage is, however, not independently recognizable, but mainly established in correspondence with the tone dot material. Tone dots are mostly used to note the pitch of individual words in isolation. While we have a fairly good idea of the basic phonetic and phonological system and the pitch shapes of a large group of words, particularly nouns, what we can say about the morphophonology of pitch, for example in the formation of inflected word forms, is quite limited. Kindaichi's prosodic classes to large extent coincide with, and were established with reference to, the distinct classes in the Ruiju-myi5gi-shi5. This has led to an unfortunate identification of the prosodic system of eleventh-century Kyoto with that ofpJ (which is obviously much older), and further to the idea that the prosodic systems in modem dialects, sometimes even including Ryukyuan, all somehow derive from eleventh-century Kyoto. These views to some extent still inform research on the historical prosody of Japanese. However, as might be expected, subsequent research has shown that the dialectal correspondences are more complicated, especially when material from Ryukyuan is taken into consideration, and that more classes must be reconstructed for pJ than are in Kindaichi's system or in the Ruiju-myi5gi-shi5. Furthermore, philology has made clear that the Ruiju-myi5gi-shi5 system has more classes than Kindaichi's system. Kindaichi's system is, however, so ingrained that newly discovered or established classes are designated as subclasses, here noted by letters, e.g. 1.3a. 7.4.1.1 Final falling pitch: classes 2.5, 3.5b, and 1.2 The ll: dot (or 'light'('-?£) lJL dot) is used in some copies of the Ruij"u-myogisho and a few other sources. It is mainly used to mark the final syllable of nouns belonging to class 2.5. It is not, however, used in all copies oftheRuij"umyogi-sho, nor is it used consistently for all 2.5 nouns. Membership of class 2.5 may well have expanded over time, so it is difficult to judge in all cases whether particular nouns originally belong to 2.5 or 2.4. Today the distinctive LF melody of this class is emblematic of Kansai dialects and the class is attracting members which originally belonged to other classes, as well as new words, such as demo 'demonstration', era 'erotic(ism)', saba 'sabotage'. The distinction between class 2.4 and 2.5 is only very rarely reflected outside Kansai and vicinity, with aim ost all dialects on either side having no distinction

216

7 EMJ phonology

between 2.4 and 2.5, and it has been suggested that class 2.5 split off from class 2.4 as an innovation in Kyoto which only spread in Kansai, but that is very unlikely, especially because there are no identifiable conditioning criteria for the split, and because the distinction seems to be reflected for example in Taisha-cho in Shimane prefecture (Uwano Zendo, p.c.). A small number of 3.5 nouns usually noted as ljl.ljl.1:: are in some sources noted as ljl.ljl.Jl:i:. For that reason 3.5 is divided into two subclasses, one with the melody LLH (3.5a) and one with LLF (3.5b), paralleling the final fall of class 2.5. Classes 3.5a and 3.5b merged completely soon after EMJ and the original membership of 3.5b is difficult to establish~ the words listed in (32) below seem to be those for which this shape is attested. It is finally noteworthy that the falling pitch of 1.2 nouns is not noted in isolation by a Jl:i: dot, but by a 1:: dot.

7.4.1.2 Initial rising pitch: class 1.3b; classes 2.3-5b? The :;iS dot is primarily used for the pitch of a number of one-mora nouns. They belong to class 1.3b, with rising pitch, and are distinct from 1.3a nouns which have low pitch, noted by lJL. These two subclasses merged soon after the EMJ period. In addition, there is a small number, fewer than twenty, dissyllabic nouns noted with initial :;iS, see (34), where SJ loanwords are in bold: (34) a.

b.

c.

:;lsljl. (RL): pagi 'shank' (also paagi LHL),yasi 'palm tree', madu 'first' (also HL), mosi 'perchance',yupi 'mating (of livestock)' :;lsl:: (RH): yuri 'lily', yusi 'name of a spinous evergreen', sime 'hawfinch', para (no NJ attestation) 'trumpet (hom)', tisa 'lettuce', yaya 'a bit', pemi 'snake' (also penbi LHH), kisa/kiza 'elephant', kuko 'Chinese matrimony vine' :;lsljl.- :;lsl::: pobo 'about', pime 'hawfinch', biwa 'loquat', goma 'sesame' (pitch only noted for the first mora~ could be either RH or RL)

These words which have rising pitch in the first syllable are generally treated as variants ofLL (2.3) or LH (2.4). They are sometimes interpreted as having had long vowels with a LH contour, e.g. *yuuri [LHH] 'lily' or paagi [LHL] 'shank'. The double attestation of 'shank' might be taken to support that, but the question arises whether paagi is not rather written with two vowels to accommodate a notation of a pitch contour, rather than the contour showing the presence of a long vowel. None of these forms is written in contemporary sources with long vowels or is reflected in later stages of the language with long vowels, so it is unlikely that the forms in (34) in EMJ had phonemically long vowels. It is more likely that the pitch contour reflects a much earlier,

7.4 Prosody; 'accent'

217

pJ or pre-OJ, long vowel or diphthong with a LH melody which by EMJ had contracted or shortened but retained the contour. Note, however, that yupi is a phonologically simplified SJ loanword (JQt!ft) which is also found in the shape iupin (> NJ yuuhin), supporting the possibility of a long vowel in this word. SJ loanwords are in boldface in (34); it is possible that the R pitch in these words reflects length or contour in the source words. It is thus likely that classes 2.3 and 2.4 (and possibly 2.5) should actually each be split into two classes, which had all but merged at the EMJ stage, but were originally distinguished by whether they originally had low pitch in the first syllable (2.3a, 2.4a, 2.5a) or rising pitch (2.3b, 2.4b, 2.5b). 7.4.2

Eleventh-century prosodic classes and later changes

Table 7.3 shows the prosodic classes, including subclasses, represented in the Ruiju-my8gi-sh8 (but not pJ classes not reflected in the Ruiju-my8gi-sh8) and the reflexes in cNJ Kyoto (based on Martin 1987, especially pp. 600-34), that is to say, the changes between EMJ and cNJ Kyoto. Table 7.3 also shows some basic correspondences with Tokyo and Kagoshima dialects; importantly, the latter show inter-dialectal correspondences, not developments from the EMJ system: the split between the Kyoto and Tokyo branches appears to predate by quite some time the (EMJ) written sources noting pitch shapes. Note, however, that Kagoshima class 'A' corresponds well to EMJ (Kyoto) High wordtone and Kagoshima class 'B' to EMJ Low wordtone, although the changes which took place between EMJ and cNJ Kyoto to some extent have upset these correspondences. It must also be noted that these correspondences are general tendencies and that many individual words do not conform to them. Examples of words belonging to the classes are given in (35). (35)

1.1 1.2 1.3a 1.3b 2.1

ka 'mosquito', yo 'world', ti 'blood', mi 'fruit' pa 'leaf', pi 'sun', na 'name', ya 'arrow', ne 'price' na 'weed', me 'eye', ki 'tree', ko 'basket', no 'field' pa 'tooth', me 'woman', su 'nest', mo 'garment' pana 'nose', arne 'gluten', niwa 'garden', kuti 'mouth', sake

2.2

kami 'paper', pita 'person', mura 'village', piru 'daytime', kata 'direction' pana 'flower', kami 'spirit, god', kami 'top hair', yama 'mountain', mimi 'ear' wara 'straw', ine 'riceplant', kami 'above', kesa 'this morning', ito 'thread' arne 'rain', asa 'morning', koto 'harp', yoru 'night', muko 'bridegroom'

'sake'

2.3 2.4 2.5

218

7 EMJ phonology

Table 7.3 Prosodic classes cNJ EMJ

J: J:

cNJKyoto

Tokyo

Kagoshima

H ('o) F ('6) L ('o) H ('o) L ('o) F ('6)

L-H (o) L-H (o) H-L (6) H-L (6) L-H (o)

H-L (A) H-L (A) L-H (B) L-H (B)

1.2 1.3a 1.3b

.jz

H F L R

2.1 2.2

J:J: J:'Jl.

HH HL

HH ('oo) HL ('6o)

LH-H (oo) LH-H (oo) LH-L (o6)

HL (A) HL (A)

2.3 2.4 2.5

'¥-'¥'f.ftlt

LL LH LF

HL ('6o) LH ('oo) LF ('o6)

LH-L (o6) HL-L (6o) HL-L (6o)

LH (B) LH (B) LH (B)

3.1 3.2

J:J:J: J:J:'Jl.

HHH HHL HLL LLL LLH LLF LHH LHL

LHH-H (ooo) LHH-H (ooo) LHH-L (oo6) HLL-L (6oo) LHH-L (oo6) LHH-H (ooo) HLL-L (6oo) LHH-H (ooo) LHH-H (ooo) HLL-L (6oo)

LHL (A) LHL (A)

J:'Jl.'Jl.

HHH ('ooo) LHL ('o6o) HLL ('6oo) HLL ('6oo) HLL ('6oo) (- HHL ('o6o)) HLL ('6oo) HLL ('6oo) LLH ('ooo) LHL ('o6o)

1.1

3.3 3.4 3.5a 3.5b 3.6 3.7

'¥-

'Jl.J:

'¥-'¥-'¥'Jl.'Jl.J:

'f.'f.ftlt 'Jl.J:J: 'Jl.J:'Jl.

LHL LLH LLH LLH LLH LLH

(A) (B) (B) (B) (B) (B)

(35 cant) 3. 1 katati 'shape', akari 'light', minato 'port', kodomo 'child', miyako 'capital' tokage 'lizard', aduki 'red bean', mukade 'centipede', 3.2 tubasa 'wing', enoki 'hackberry' 3.3 patati 'twenty', ikura 'how much?', komugi 'wheat', tikara 'strength', awabi 'abalone' 3.4 kagami 'mirror', pakama 'man's skirt', inori 'prayer', kotoba 'word', otoko 'man'

7.4 Prosody; 'accent'

3.5a 3.5b

3.6 3.7

7.4.3

219

iraka 'roof tile', sekido 'barrier', asuka 'morning fragrance', taira 'fiat', teboko 'spear' aoto 'bluegrey grindstone', pirome 'broad-leaf seaweed', pitoe 'single layer', tamaki 'elbow guard', akidu 'dragonfly', pimidu 'ice water' usagi 'rabbit', padaka 'naked', kamome 'seagull', nezumi 'rat', sumomo 'plum' kabuto 'helmet', tamago 'egg', kusuri 'drug', sukosi 'a little', mabuta 'eyelid'

Phonological interpretation

As mentioned above, work on historical Japanese prosody is often done in tenns of phonetic pitch shapes or patterns, but there is no standard or agreed description of the phonological categories of which the pitch patterns are manifestations. Unsurprisingly, EMJ appears similar to modem Kyoto (see Table 7.3 above) which is its direct descendant, but it is not possible to describe all the distinctive pitch shapes exhaustively in terms of wordtone and accent, without recourse to multiple accents within a word, or more than two levels of wordtone, so we need something else, or something more. However, it is clear that wordtone, the initial pitch of a fonn, is a significant lexical property, also at the EMJ stage. This is reflected in a simple, but important, general rule for the prosodic shape of complex fonns which seems to have held through the history of the language, given in (36). See (39) below for a number of examples involving compounds. (36)

A complex fonn has the wordtone of its initial component.

This rule, which is sometimes referred to as 'Kindaichi's Law' (after Kindaichi Haruhiko), applies to the formation of all complex forms, including prefixation, compounding, and fonnation of inflected fonns. Because of this rule, wordtone is often used as a diagnostic in etymological studies, such that two words are thought not to be derivationally related, and therefore not cognate, if their wordtone, as reflected in the earliest materials, is different. However, although wordtone is a basic lexical property it can, like all other phonological features, change in the course of conditioned regular sound change, as shown by the changes in Table 7.3, where for example classes 2.3, 3.2, 3.4, and 3.5 today have a different wordtone from EMJ. This type of change has resulted in many attested cases of tonal divorce of related fonns, as exemplified in (37) where the wordtone of a lexicalized compound has changed in the course of regular sound change and is now different from that of its original first component. This type of change must also be envisaged before the attestation of

220

7 EMJ phonology

prosody in the written sources, showing that it is dangerous to rely too heavily on tonal correlations in etymology. Opinions differ about whether wordtone as such should be reconstructed for pJ, or whether it reflects one or more segmental features of pJ, such as for example initial voiced consonants or vowel length, but it is clear that wordtone is a significant lexical property in Japanese whose history goes much further back than the eleventh-century sources. (37)

kage (- kaga-) 'light, shade, reflection' kagami 'mirror'(< kaga- + mi- 'see')

EMJ

NJKyoto

LF

LF

LLL

HLL

Wordtone alone can account for the shape of those prosodic classes whose overall phonetic melodies are level (with no drop in pitch), either all [high] (1.1, 2.1, 3.1) or all [low] (1.3a, 2.3, 3.4), and interestingly these classes have more members than the other classes. Thus 2.1 and 2.3 together hold 59 per cent of all two-mora nouns, and 3.1 and 3.4 hold 66 per cent of three-mora nouns (Martin 1987: 273). Note also that, with a small group of exceptions, the only lexically relevant prosodic feature for verbs and adjectives is wordtone. The prosodic classes with [falling] (1.2, 2.2, 3.2, 3.3) or [rising] (1.3b, 2.4, 3.5a, 3.6) melodies can be accounted for by the presence of a single change in pitch. The identification of change in pitch (rather than specifically a fall or rise) as a distinctive phonological property is due to Samuel E. Martin (1987), and we here follow his analysis and refer for EMJ to the locus of pitch change as 'accent'. Thus, the difference between HLL (3.3) and LHH (3.6) is the wordtone: both have an accent, which changes the pitch, between the second and third moras. Noting wordtone as above, but accent by a small vertical line, before the change in pitch, we can note these forms phonologically as /'o'oo/ and to'oo/. Apart from the more abstract notion of accent, this system is like that of Modem Kyoto. However, it cannot account for those classes which have a [rising-falling] melody (2.5, 3.5b, 3.7), i.e., L-words which also have a drop in pitch. One way of accounting for such forms is to posit an additional word internal Low tone. Noting this L-tone by a grave over the mora which carries it, we can show the difference between LHH (3.6) and LHL (3. 7) as t o'oo/ versus to'ool. Table 7.4 shows this interpretation applied to the attested pitch patterns for one- to six-mora nouns. On this analysis, the structural changes between EMJ and NJ Kyoto consisted in (a) the loss of the free L-tone and (b) the change of accent from marking a change in pitch to marking a fall in pitch. These are not great changes and comparing EMJ with NJ Kyoto (cf. Table 7.3 above) shows that the phonetic changes are not great either.

7.4 Prosody; 'accent'

221

Table 7.4 Phonological interpretation of the EMJ pitch patterns One-mora nouns 1.1

1.2 1.3a 1.3b

H F L R

'o ''o (or 'o) 'o ''o

Two-mora nouns 2.1 HH 2.2 HL 2.3 LL 2.4 LH 2.5 LF

'oo 'o'o (or 'oo) 'oo 'o'o 'o'o

Three-mora nouns 3.1 HHH 3.2 HHL 3.3 HLL 3.4 LLL 3.5a LLH 3.5b LLF 3.6 LHH 3.7 LHL

'ooo 'oo'o 'o'oo 'ooo 'oo'o 'oo'o 'o'oo 'o'oo

Four-mora 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10 4.11

'oooo 'ooo'o 'oo'oo 'o'ooo 'oooo 'ooo'o 'oo'oo 'o'ooo 'o'ooo 'o'ooo 'oo'oo

nouns

HHHH HHHL HHLL HLLL LLLL LLLH LLHH LHHH LHHL LHLL LLHL

Five-mora 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12 5.14 5.15 5.16

nouns

HHHHH HHHHL HHHLL HHLLL HLLLL LLLLL LLLLH LLLHH LLHHH LHHHH LHHHL LHHLL LLHHL LLHLL LLLHL

Six-mora nouns 6.1 HHHHHH 6.2 HHHHHL 6.3 HHHHLL 6.7 LLLLLL 6.10 LLLHHH 6.11 LLHHHH 6.13 LHHHHL 6.14 LHHHLL 6.17 LLHHHL 6.20 LLLHHL 6.21 LLLHLL 6.22 LLLLHL 6.23 LLLLLH

'ooooo 'oooo'o 'ooo'oo 'oo'ooo 'o'oooo 'ooooo 'oooo'o 'ooo'oo 'oo'ooo 'o'oooo 'o'oooo 'o'oooo 'oo'ooo 'oo'ooo 'ooo'oo 'oooooo 'ooooo'o 'oooo'oo 'oooooo 'ooo'ooo 'oo'oooo 'o'ooooo 'o'ooooo 'oo'oooo 'ooo'ooo 'ooo'ooo 'oooo'oo 'ooooo'o

Needless to say, other analyses are possible and have been proposed. Regardless of the specifics, it is clear that EMJ had wordtone as a distinctive lexical category, and that at least one more tone is required. On the analysis proposed here, the free L-tone is severely restricted: it occurs only after an accent in L-words and always on the final or penultimate mora. It is quite possible that the pre-EMJ system had more tones in freer combination, cf. the initial rising tone attested in a small group of two-mora nouns (7.4.1.2). It is possible to describe EMJ entirely as a restricted tone language, with more tones, but with no reference to accent. It would also be possible to view H-words with a fall as having the free L-tone, rather than an accent, so that

222

7 EMJ phonology

for example HLL is analysed as /'ooo/ instead of /'o'oo/, leaving the category of accent relevant only in L-words. However, the EMJ system developed into that of modern Kyoto with wordtone and accent, so the free L-tone, and any other tones posited in alternative analyses, were lost in any case. It is also clear that accent in subsequent periods has been a relevant phonological category for both H- and L-words. The question of when accent was introduced in the language (if it was not always there), could be seen to depend on theoretical definitions of 'accent'. More importantly for the analysis ofEMJ, however, is the fact that what little we can say about the prosodic shape of complex forms seems to involve the notion of accent as defined by :Martin (as the locus of change in pitch, rather than specifically a rise or fall).

7.4.4

Complexforms

An important part of the prosody of NJ is the morphophonology of wordtone and accent, in composition, in the combination of lexical words and particles, and in the formation of inflected word forms. Unfortunately, material with tone dots mostly gives information about lexical words in isolation, so apart from composition we know little in detail about this for EMJ.

7.4.4.1 Noun + noun compounds Noun+ noun (N1 + N 2) compounds, like other complex forms, have the wordtone of the initial morpheme, that is ofN 1. As a general rule, compounds except under the conditions stated in (38b) have an accent, a change of pitch, usually either between the penultimate and final mora, or between the ante-penultimate and penultimate mora (in which latter case it is usually followed by a free L-tone); (38a) is the general rule, but spells out the circumstances in which the compound has an accent. (38) a.

A compound has an accent (if it has five or more moras or ifN2 has an accent) b. unless it has four moras or fewer and N 2 has no accent, in which case the compound has no accent.

The regularities in (38) hold reasonably well, but there are a fair number of exceptions. A few examples are given in (39), ordered by the rule determining whether the form has an accent and secondarily by the wordtone ofN 1. (39) a.

kawa HL 'river'+ yanagi HHH 'willow'=> pito HL 'person'+ kasira LLL 'head' =>

kawayanagi HHHHL 'oooo'o 'purple willow' pitogasira HHHHL 'oooo'o 'weathered skull'

223

7.4 Prosody; 'accent'

pana HH 'nose'+ pasira LLH 'pillar'=> kami HL 'paper'+ zeni LH 'money'=> turi HH 'fishing' + pune LH 'boat'=>

panabasira HHHHL 'oooo'o 'bridge of the nose' kamizeni HHHL 'ooo'o 'paper money' turibune HHHL 'ooo'o 'fishing boat'

kasira LLL 'head'+ kasa HH 'sore'=> mimi LL 'ear'+ kusari HHH 'chain' => yama LL 'mountain'+ ubaraHHL 'rosebush' => pama LL 'beach'+ kuri HL 'chestnut' => kinu LH 'silk' + kasa LH 'umbrella'=> tate LF 'vertical'+ isi HL 'stone'=>

kasiragasa LLLHL 'ooo'oo 'scalp sore(s)' mimigusari LLLHL 'ooo'oo 'ear pendant' yamaubara LLLHL 'ooo'oo 'wild rose'

b. tori HH 'bird' + ami LL 'net'=> kasa HH 'sore' + puta HH 'lid'=> wara LH 'straw' + puta HH 'lid'=> mayu LF 'eyebrow'+ sumi LL 'ink, charcoal' => mugi LH 'barley' + nawa LL 'rope'=>

pamaguri LLLH 'ooo'o 'clam' kinugasa LLLH 'ooo'o 'silk umbrella' tateisi LLLH 'ooo'o 'upright stone' tor"iami HHHH 'oooo 'bird net' kasabuta HHHH 'oooo 'scab' warabuta LLLL 'oooo 'straw lid' mayuzumi LLLL 'oooo 'eyebrow paint' muginawa LLLL 'oooo 'cruller'

7.4.4.2 Particles Judging from the very limited material which notes the pitch of short phrases, it is usually assumed that many grammatical particles had their own independent prosodic shapes in EMJ, contrasting with later stages of the language where many particles are prosodically neutral. Thus EMJ particles such as ga (genitive), wo (accusative), wa (topic), and kara (ablative) seem to have been H or HH. Genitive no is thought to have been prosodically neutral, like NJ ga. On the other hand, to (comitative), mo (emphatic apic), andyori (ablative) seem to have been 'pre-accenting', that is, to have begun on the opposite pitch of the immediately preceding mora, so that yori had the shape 'oo, e.g.

224

7 EMJ phonology

moto-yori LLHL ('oo'oo) 'originally' (moto LL 'root, original'), but kuti-yori HHLL ('oo'oo) 'from the mouth' (kuti HH 'mouth'). Overall, the material is inconclusive in most matters of detail. 7.4.4.3 Verbs and a4Jectives Verbs and adjectives are mainly lexically specified for wordtone. This has held through all attested stages of the language and sets the inflecting, predicative word classes, verbs and adjectives, off from the uninft.ecting nouns. As in the modern language, it is thought that different inflected forms in EMJ were assigned different prosodic shapes by rule. Adjectives fall in two prosodic classes distinguished by wordtone. However, the distinction is not found among one-mora stems, all four of which (yo' good', na- 'non-existent, lacking', su- 'sour', to- 'sharp') are L (ko- 'thick' and u- 'lamentable, sad' do not have safe attestations of their prosodic class). Adjective stems of two or more moras distinguish H and L wordtone, e.g. H atu- 'thick', aka- 'red', atarasi- 'precious', kaUbasi- 'fragrant', but L atu- 'hot', taka- 'high',pagesi- 'severe', omosiro- 'amusing'. That distinction largely later merged also for stems of two or more moras. Too little is known about the prosodic shapes of inflected forms to say anything firm. For verbs, the relevant prosodic classes are H- and L-verbs, with a further subdivision in terms of accent only among three-moraL-verbs. The 'prosodic length' of a verb is a relevant lexical property which for consonant base verbs is measured on the basis of its infinitive or conclusive form, not its basic stem, hence notations like ok(V)-. The class of accented L-verbs of three or more moras is not large; they are usually thought to have originated in compounds. The basic verb paradigms in Table 7.5 are based on Martin (1987: 191-8). In the sources, verbs are generally listed in the conclusive form. Other forms are rarely represented and the evidence is conflicting. The forms posited here are to some extent hypothetical and differ in some respects from other accounts, reflecting that much is still unclear about the prosodic shape of inflected forms. Importantly, for example, for two-moraL-verbs and three-mora class 1 L-verbs Kindaichi and Wada (1980) give the conclusive form as LF and LLF and the adnominal form as LH and LLH. It must also be noted that the prosodic shape of the adnominal form is based on fairly small numbers of examples of phrases in which a noun is modified by an adnominal form, and it is not possible to know whether the adnominal used in other contexts had the same prosodic shape. The overall regularities governing the paradigms in Table 7.5 are that the conclusive and the infinitive have an accent (change of pitch) between the penultimate and final mora (or, in the case ofmono-moraic forms, within that mora), whereas the adnominal does not have an accent. The exception to this rule are the lexically accented 3-mora L-verbs (Class 2), which have a Low

225

7.4 Prosody; 'accent' Table 7.5 Prosodic shapes ofEM.! basic inflected verb forms One-mora H-verb

One-moraL-verb

ki'wear'

se'do'

mi'see'

lw'come'

pe-

kiru HL

su

"o'o

FIH Ao I 'o

miru LH 'o'o

ku R ''o

pu R ''o

Adnominal

kiru HH 'oo

suru HH 'oo

miru LL 'oo

kuru LL 'oo

puru LL 'oo

Infinitive

ki

si

FIH Ao I 'o

mi R ''o

ki R ''o

pe

FIH Ao I 'o

Conclusive

1\vo-mora H-verb

R ''o

1\vo-mora L-verb

ok{V)'put'

age'raise'

puk(V)'blow'

puke 'deepen'

oku HL

agu HL

"o'o

"o'o

puku LH 'o'o

puku LH 'o•o

Adnominal

oku HH 'oo

aguru HHH 'ooo

puku LL 'oo

pukuru LLL 'ooo

Infinitive

old HL

age HL

'o'o

'o'o

puki LH 'o'o

puke LH 'o•o

Conclusive

'pass'

Three-mora H-verb

Three-moraL-verb

sukum'crouch'

padime'begin (vt.)'

Class 1 (unaccented) tutum{V)awase'bundle' 'join (vt.)'

Class 2 (accented) aruk{V)kegare'walk' 'get soiled'

Conclusive

sukumu HHL 'oo•o

padimu HHL 'oo•o

tutumu LLH 'oo•o

awasu LLH 'oo'o

aruku LHL 'o'oO

kegaru LHL 'o•oo

Adnominal

sukumu HHH 'ooo

padimuru HHHH 'oooo

tutumu LLL 'ooo

awasuru LLLL 0000

aruku LHH 'o'oo

kegaruru LHHH 'o'ooo

Infinitive

sukumi HHL 'oo•o

padime HHL 'oo•o

tutumi LLH 'oo•o

awase LLH 'oo'o

aruki LHL 'o'oO

kegare LHL 'o•oo

226

7 EMJ phonology

tone after the lexical accent in the conclusive and infinitive, but not the adnominal. The substantive noun is segmentally identical with the infinitive, but has no accent: kobi HH 'flattery' (kobi- (H) 'flatter'), ami LL 'knitting' (am- (L) 'knit'), kazari HHH 'ornament' (kazar- (H) 'adorn'), urami LLL 'resentment' (uram- (L) 'resent'). The gerund is attested in a number of cases and generally seems to have attached the gerund formant -te with an invariable H pitch to the infinitive, e.g. oite HLH (ok- 'put'), konde LHH (kom- 'enter'), padimete HHLH (padime- 'begin (tr.)'), agaite LLHH (agak- 'paw'), but there are already instances of -te attaching with the same pitch as the preceding mora, e.g. oite HLL (ok- 'put'). REFERENCES EMJ phonology: Frellesvig 1995, Mabuchi 1971, Martin 1987, Wenck 1959. Onhin: Frellesvig 1995, Komatsu 1981: 16lff. On the development of OJ /p/: Kiyose 1985. Renjo: Vance 1987: 164-7. Accent: Frellesvig 1998, Kindaichi and Wada 1980: 7-10, Martin 1987, Matsumori 2008, Shimabukuro 2007, Uwano 1977.

8

Grammar

8.1

Verbs

8.1.1

Morphological categories

The overall range of inflected categories and fonns is not very different between OJ (cf. 3.1.3) and EMJ; the EMJ inventory of fonns is shown in Table 8.1. The greatest change was the loss of the nominal fonn, e.g. kakaku, akuraku, which was prominent in OJ (see 3.1.3.3). It went out of general use, but was archaically retained in kanbun-kundoku and to some extent in poetry; it also survived in some set expressions, cf. 9.1.7. The OJ uses of the nominal fonn came to be taken over (a) by nominalizations by nouns such as koto 'thing' which increased in use both in abstract nominalizations ('the fact that ... ') and in complement clauses, and (b) by the adnominal form which expanded its use in headless nominalizations; see further 12.6.1 about changes in the use of the adnominal fonn. The prohibitive retained only the one pattern shown in Table 8.1, and of the several related optative verb fonns, only that in -(a)namu survived in EMJ, with the general sense 'I wish'. A new optative fonn in-(a)baya came into use, e.g. kikabaya 'I wish to hear', originating in the combination of conditional -(a)ba and the particle ya. The segmental distinction between exclamatory and imperative, e.g. OJ kake versus kakye, was lost, in the course of the merger of OJ Ce and Cye (7.3.2.1). For vowel base verbs, the imperative now generally ended in -yo (which had been optional with some verb classes in OJ, cf. 3.4.5).

8.1.2

Conjugation classes and basic paradigms

In the first half ofEMJ, the so-called shimo ichidan (lower monograde, abbreviated as 'LM') conjugation class appeared, morphophonologically patterning with UM., but with a different stem vowel. The basic paradigm of the EMJ vowel base verbs is given in Table 8.2. Through EMJ and part of LMJ the LM verb class only had a single mem her, namely 'to kick'. It is thought that this verb had the base shape /kwe/, rather 227

228

8 EMJ granunar Table 8.1 EM/ inflected verb forms

QD

LB

Base

kale-

alee-

Finite Conclusive Adnominal Exclamatory Imperative Negative conjectural Optative Optative Prohibitive

kaku kaku kake kake kakazi kakabaya kakanamu nakaki so

aku akuru akure aleeyo aleezi aleebaya aleenamu na alee so

Non-finite Infinitive Gerund Continuative Conditional Provisional Concessive

kaki kaite - kakite kakitutu kakaba kakeba kakedo(mo)

alee aleete aleetutu aleeba akureba akuredo(mo)

Table 8.2 Basic parodigm ofEMJ vowel base verbs LM

LB

UM

UB

s-irr

k-irr

Base

kwe'kick'

alee'open, tr.'

mi'see'

oki'arise'

se'do'

ko'come'

Infinitive Conclusive Adnominal Exclamatory Imperative

kwe kweru kweru kwere kweyo

alee aku akuru akure aleeyo

mi miru miru mire miyo

oki oku okuru okure okiyo

si su suru sure seyo

ki ku kuru kure koyo

than just Ike/. The evidence is not incontrovertible and this is in that case the only native word with /kwe/. 1 EMJ kwe- reflects the OJ LB verb kuwe-. This change could be seen as an early beginning of the sweeping change of all bigrade verbs to monograde conjugation which in the main took place in NJ (15.1.2.1), but it is intriguing that this singular verb, which was not particularly frequent or otherwise prominent, changed in this way and maintained 1

The evidence includes writings such as 11::::0 which is interpreted as kweru rather than keru because 11::: had the SJ (go-on) reading lk:we/.

229

8.1 Verbs Table 8.3 Basic paradigm of EM.! CVi- and CVe- base verbs UB

LB

LB

LB

Base

kui- I/kuyilI 'regret'

OJ taye > toe- 1/taye/1 'cease (intr.)'

OJ kape- > kawe- > kae- llkaell

'exchange'

uwe- > ue Ifuel I 'starve'

Infinitive Conclusive Adnominal Exclamatory Imperative

kui kuyu kuyuru kuyure kuiyo

toe tayu tayuru tayure toeyo

kae kau kauru kaure kaeyo

ue uu uuru uure ueyo

its separate conjugation for a relatively long period. It may be its unique stem shape which contributed to the stability of the conjugation. Another early beginning of the change of bigrade to monograde verbs is the migration of one-mora UB verbs to UM, which was under way in OJ and was completed in EMJ (cf. 3.4.3.1). Combined with the merger of Ci and Cwi (7.3.2.1) this resulted in a clear-cut distribution of -i- base verbs between UM and UB such that all one-mora -i base verbs were UM and all -i base of more than two moras verbs were UB. Despite the emergence of the LM class, a similar lengthbased distribution did not arise for the-e base verbs, as all other one-mora-e base verbs than kwe- (e.g. e- 'get', pe- 'pass') remained LB. Several sound changes gave LB verbs of the shape CVe- (loss of /y/, 1-p-1 > 1-w-1 and loss of 1-w-1; cf. 7.3.2.5). Verbs which reflected OJ C~e- remained //Cvye/lin underlying representation, just like UB CViverbs (3.4.1.2), whereas those which reflected OJ CVpe- and CVwe- became 1/CVe// also in underlying representation. This is shown by the different conclusive, adnominal and exclamatory forms of these verbs, see Table 8.3. In LMJ all CVe- verbs became //Cvye// underlyingly (see 12.3.1). 8.1.3

Verbal nouns

The class of verbal nouns, predicated by se- 'do', originates in EMJ, and in particular in the large intake of SJ loanwords, all of which were taken in morphologically as nouns. Some early examples are given later in 9.2.3.2. In OJ one use of se- was as a light verb in constructions with a verb infinitive (see 3.4.3.2.1), and the emergence of the verbal nouns may be seen as an extension of this use. Se- is still used as a light verb in NJ. After moraic nasals Ji, U, N/, se- became ze- (cf. 7.1.2.2). By far the greatest number of verbal nouns are SJ loanwords, but a few native forms are also included, especially verbs resulting from the derivational pattern ADJ-nze- reflecting the

230

8 EMJ granunar

OJ construction ADJ-mi se-, 'ADJ-ACOP.INF do; make ADJ of, take ADJ-ly', e.g. karo-mi se- > EMJ karo-nze- > NJ karonji- 'make light of, take lightly' (karo'light'), omo-mi se- > omo-nze- > omonji- 'value' (omo- 'heavy'). 8.1.4

Consonant base verbs

The basic paradigms of the consonant base verbs are shown in Table 8.4. In the course of /-p-/ > 1-w-1, OJ -p bases became 1-w-1 bases around 1000. As /w/ was lost before /e/ around 1100, the exclamatory and imperative changed so that the base consonant, as in NJ, only appears before /a/. The onbin sound changes (7.1.4) affected the infinitive of -p, -k, -b, -g, -m, -n base verbs, ending in -pi, -ki, -bi, -gi, -mi, -ni, resulting in the emergence ofvariants of the infinitive when used as a stem for suffixation, generally referred to as the onbinkei. This variation was analogically extended also to consonant base verbs whose infinitives were not directly involved in the onbin sound changes, that is -t, -s, -r bases whose infinitives ended in -ti, -si, -ri, so that all consonant base verbs acquired a new conditioned, bound stem used for suffixation of some auxiliaries and inflectional endings. With -m, -b, -p bases, both vocalic and consonantal onbin stems resulted. These variants belonged to different varieties, but little concrete is known about their distribution. The consonantal variants are thought to be more typical of eastern dialects; the distribution in modern dialects corresponds largely with the major east-west dialect boundary, with vocalic variants prevalent on the western side and consonantal variants on the eastern side. The consonantal variants also appear later in the written sources than the vocalic ones. This is sometimes taken to reflect that the consonantal variants developed from the vocalic ones, but it is more likely due to a later and/or lesser representation in the written sources of the varieties in which they arose and were used. The onbin stem was used with a variety of suffixes. As in NJ, a major use of the onbin stem was from early on with the gerund formant -(i)te, and it was also used regularly with the stative auxiliary (8.4.2) -(i)tar- (which originates in the OJ periphrastic construction -(i)te ar-), and with the respect auxiliary verb -tamap- (> -tamaw-). The initial /t/ of these suffixes changed to /d/ after stem-final nasals Jl, U, N/ (7.1.2.2). Rules for the distribution of the onbin stem must have differed between speakers and also changed over time. In EMJ onbin stems were also used before, for example, -si (adnominal of simple past -(i)ki) and somewhat more rarely before perfective -(i)te- and -(i)n-, modal past -(i)ker-, and the conclusive of simple past -(i)ki. In LMJ, perfective -(i)te- was regularly used with the onbin stem (until -(i)te- went out of use, see 12.1.3.1 ), and a general feature is that the suffixes regularly selecting the onbin stem have initial t-. However, the LMJ desiderative auxiliary -(i)ta(12.1. 5 .3) was not used with the onbin stem and today reflexes of -tamaw- do

8.1 Verbs

231

Table 8.4 Basic paradigm of EM.! consonant base verbs n-irr

r-irr

QD

QD

Base

sin-

ar-

tor'take'

mot'hold'

a- stem onbin stem

sina siN-

ara aQ-

loratoQ-

motamoQ-

Infinitive Conclusive Adnominal Exclamatory Imperative

sini sinu sinuru sinure sine

ari ari aru are are

tori toru toru tore tore

moti motu motu mote mote

QD

QD

QD

Base

kale'write'

kog'row'

sas'pierce'

a- stem onbin stem

kakakai-

kogako"i-

sasasai-

Infinitive Conclusive Adnominal Exclamatory Imperative

kaJci kaku kaku kake kake

kogi kogu kogu koge koge

sasi sasu sasu sase sase

QD

QD

QD

QD

QD

pre-1000

1000-llOO

post-llOO

Base

yom'read'

yob'call'

kap- > 'exchange'

kaw-

kaw-

a- stem onbin stem

yornayou- -yoN-

yobayou-- yoN-

kapakau- -kaQ-

kawakau- -kaQ-

kawakau- -kaQ-

Infinitive Conclusive Adnominal Exclamatory Imperative

yomi yomu yomu yome yome

yobi yobu yobu yobe yobe

kapi kapu kapu kape kape

kawi kau kau kawe kawe

kai kau kau kae kae

not select the onbin stem. In writing, onbin stems did not replace the infinitive in combination with these suffixes, but coexisted with it. To some extent this probably reflects variation in use, partly in rules for the distribution of the onbin stems, especially with those suffixes which were infrequent with the onbin stem, and partly in stylistic value. On the other hand, this is also thought to reflect norms which inhibited the use of onbin forms in writing; it is likely

232

8 EMJ granunar

that these fonns were used much more widely in speech, as suggested for example by their frequent use in kunten materials. Onbin stems never became part of the classical literary nonn, and they are usually not represented in the Classical Japanese verb paradigms in the school grammar. When noting inflectional endings or auxiliaries which regularly attach to the onbin stem, we will use the prefix '(!)', similarly to the '(a)', '(i)', and '(e)' which show that endings attach to the a stem, the infinitive, or the exclamatory (see 3.4.4). Thus, we will write -(I)te gerund fonnant, -(!)tar- stative auxiliary, but not *-(I)ker-, because the modal past does not seem to have been used as regularly with the onbin stem. As with the other morphophonological prefixes, the '(I)' in -(I)te or -(!)tar- is not part of the phonemic shape of the suffixes, but merely shows that the formant attaches to the onbin stem. In addition to the regularized new system of consonant base verb stems, the adnominal fonn of r-irr verbs or suffixes belonging to the r-irr conjugations generally assimilated to a following nasal or media in a grammatical morpheme, especially extensions such as be- necessitive, nar- hearsay, mer- evidential, e.g. naru-besi 'coP-NEC' => naNbesi, paberu-meri 'exist.HUM-EVID' => pabeNmeri, aru-nari 'exist-hearsay' => aNnari.

8.2

Adjectives and copula

In the course of EMJ both the regular and the adjectival copula acquired secondary, full verbal conjugations (8.2.1), while also retaining their primary paradigms. The primary paradigm for the regular copula was highly defective, and that for the adjectival copula lacked a num her of the categories which verbs inflected for. The primary forms of the adjectival copula underwent a great deal of change between OJ and EMJ. The forms in common use in EMJ are listed in Table 8.5, contrasting with those used in OJ shown above in 3.2.1. Apart from simplification, in the sense of elimination of variant fonns for the same categories, the nominal form in -kyeku, which was frequent in OJ, was lost. Two formatives, infinitive-2 -mi and exclamatory-2 -sa, changed to become derivatives, as in NJ where they derive concrete and abstract nouns from adjectives. -mi is also reflected in the derivational pattern ADJ-nze mentioned above (8.1.3). Also the adjectives acquired new variant fonns as a result of the onbin sound changes (7.1.4): adnominal OJ -ki > EMJ -i (- -ki) and infinitive OJ -ku > -u (- -ku). These fonns were not used in the poetry, but are found scattered throughout the prose texts. As with the verbs, their actual use in speech was probably somewhat wider than their textual occurrence suggests. As opposed to the verbal onbin stem, the adjectival onbin forms were not subject to grammatical conditioning.

233

8.2 Adjectives and copula Table 8.5 EM/ adjectival copulaforms Base

taka' tall'

asi'bad'

Finite Conclusive Adnominal Exclamatory

si ki- i kere

taka-si taka-lei - taka-i taka-kere

asi asi-ki - asi-i asi-kere

Non-finite Infinitive Gerund Conditional Provisional Concessive

leu- u /cute- ute kupa> kuwa kereba keredo(mo)

taka-leu- taka-u taka-/cute - taka-ute taka-kupa > taka-kuwa taka-kereba taka-keredo(mo)

asi-ku - asi-u asi-kute - asi-ute asi-kupa > asi-kuwa asi-kereba asi-keredo(mo)

Table 8.6 EM/ copula forms Adnominal Infinitive Gerund

no ni nile > de; nisite

to to site

The very defective EMJ primary copula paradigms are shown in Table 8.6; cf. 3.3.1 which has a few more forms. A new gerund, de, is found from the end of the period, originating in contraction of nite.

8.2.1

Secondary conjugations: ar- extendedforms

In OJ the existential verb ar- was used with the infinitive of the adjectival copula (-ku), the regular copula(ni, to) and the negative auxiliary (-zu) to form analytic forms, and these combinations sometimes fused phonologically: ADJ-ku ar- => ADJ-kar-, NOUN-ni ar- => NOUN-nar-, NOUN-to ar- =>NOUN-tar-, VERB-ZU ar=> VERB-zar- (3.4.2.1-2). In EMJ the fused forms gave rise to secondary conjugations, all of which belong to the r-irr conjugation and have the basic paradigms in Table 8.7. Note, however, that the secondary conjugations did not replace the analytic forms which have continued in use through the history of the language (cf. 12.2.3).

8.2.2

Aqjectival copula and negative

With the adjectival copula and the negative, the secondary conjugations supplemented the primary, basic paradigms and made morphophonologically

234

8 EMJ granunar Table 8. 7 Basic paradigm ofEMJ secondary conjugations of the copula, the adjectival copula and the negative auxiliary Copula

Adjectival copula

Negative

Base

nor-

tar-

-kar-

-(a)zar-

a- stem onbin stem

naronaQ-

tarataQ-

kara IraQ-

zarazaQ-

Infinitive Conclusive Adnominal Exclamatory Imperative

nari nari naru nare nare

tori tori taru tare tare

kari kari karu /rare /rare

zari zari zaru

zare zare

possible the fonnation of more inflected forms, for example an imperative, and free combination with tense and mood auxiliaries which did not combine with the primary fonns. Frequent combinations (often with multiple occurrences of fonns which incorporate ar-) include: taka-kara-zu 'isn't tall', takakara-zari-ki 'wasn't tall', taka-kara-zari-keri 'wasn't tall!', taka-kara-mu 'is probably tall', taka-kara-zara-mu 'probably isn't tall'. The main function of the secondary conjugations was combination with inflectional endings or auxiliaries, for which the a- stem, the onbin stem, and the infinitive were used, but although both adjectives and the negative auxiliary had a primary conclusive, adnominal and exclamatory (and the inflected forms built on the exclamatory), the secondary conclusive, adnominal and exclamatory were also used, although less frequently than the primary fonns. The secondary conjugation of the adjectival copula was also used with the extension be- necessitive, be-kar-, used in combination with auxiliaries such as negative, simple past, or modal past, e.g. kurabu be-kara-zari-keri 'shouldn't compare!'. A further development from be-kar- is the extension bekasi- necessitive, where bek-, or beka-, was treated as a verbal stem or root from which a shiku adjective was derived. 8.2.3

Copula

For the regular copulas, the new secondary forms became fully inflected copula fonns which complemented the defective basic paradigm, not just to form extended combinations, but for simple predications: NPl NP2-nari 'NPl is NP2'. The basic forms in Table 8.6 continued in use, however, as adverbial and adnominal copula fonns. Of the inflected copulas, tar- was restricted and mainly used with a subset of SJ adjectival nouns (8.3), as well as in kanbun-

8.3 Adjectival nmms

235

kundoku, whereas nar- became the general copula. Like the secondary conjugations of the adjectival copula and negative auxiliary, the inflected copulas continued to coexist with the analytic forms: NP-ni ari 'is NP'. Ni ar- had the variant nite ar-, with ar- following the gerund instead of the infinitive. In the course of EMJ nite ar- became de ar-, which is the source of the cNJ copula da (cf. 12.2.2, 15.2, 16.2).

8.3

Adjectival nouns

The so-called a4Jectival nouns (keiyodoshi ID$jf]~irJ, lit. 'adjectival verbs'), including words such as siduka 'calm, relaxed, quiet', kiyora 'clear', yaparaka (> yawaraka) 'soft', are usually regarded as having emerged as a separate word class in EMJ. The term 'adjectival noun' is widely accepted, and is adopted here, but 'nominal adjective' would be better, for these words are really a subclass of adjectives. Neither adjectives nor adjectival nouns occur with case particles. Whereas adjectives are adverbialized, adnominalized and predicated by the (restricted) adjectival copula, the adjectival nouns use forms of the regular copula in those functions, e.g. siduka-nari 'is calm', siduka-ni 'calmly', siduka-narazu 'isn't calm'. Many of the EMJ adjectival nouns go back to OJ and many include derivational elements which are found already in OJ, especially -ka, -raka, -yaka. Usually, forms reflected as EMJ adjectival nouns are classified as adverbs in OJ. However, this seems to reflect only that they did not occur with case particles, but these forms do not conform to usual definitions of adverb in Japanese, which include adverbial use without a copula or other particle as a criterion, e.g. ima 'now'. These forms were in OJ adverbialized by ni (or to), adnominalized by no, and predicated by ni (or to)+ existential verb, i.e. by combining with the OJ copula forms. Descriptively, therefore, there seems to be little reason not to recognize adjectival nouns as a distinct word class already in OJ: the emergence of an inflected copula (nar-) in EMJ seems to have been the impetus for recognizing this group of words as a separate word class from EMJ. The intake of a number of SJ loanwords as adjectival nouns may also have been a contributing factor, although the number of SJ adjectival nouns was not great in the first half of EMJ. Since then, however, the class of adjectival noun has been open, freely accepting loanwords (as opposed to the class of adjectives), so the vast majority of adjectival loanwords are borrowed as adjectival nouns (e.g. cNJ yunzku- 'unique' < English unique). Thus most adjectival nouns are today SJ or western loanwords. A small number of SJ adjectival nouns used the rarer copula variant tar- (see 8.2.1), e.g. daildail-tar- 'grand'. Several new EMJ derivatives derive adjectival nouns, e.g. -gati- 'be prone, likely', -ge- 'seem, appear', -yail- '-like'.

236

8 EMJ granunar

8.4

Auxiliaries

A num her of the OJ auxiliaries went out of use or changed function in EMJ, whereas other new auxiliaries emerged. The OJ respect auxiliary -(a)s- was lost, but the OJ causative -(a)sime-, which was lost as a causative, was retained in its use to express respect. See 12.7 about honorific language. The OJ passive -(a)ye- was lost, but two new causative (-sase-) and passive (-rare-) auxiliaries appear which are still in use today (8.4.1). In the tense and aspect system a number of changes took place, including the appearance of a new stative auxiliary, -{i)tar- (8.4.2). Also a number of new modal auxiliaries are found in EMJ (8.5). Auxiliaries in productive use in EMJ are shown in (1): (1)

Respect Causative Passive Negative Stative Stative Perfective Perfective Simple past Modal past Conjectural Past conjectural Intentional Optative Negative optative Subjunctive

-(a)sime-sase-rare-(a)zu - -(a)zar-(i)tar-er-(i)te-(i)n-{i)ki-(i)ker-(a)m-{i)kem-(a)nze-1-(a}flze-(a)maposi-(a)mau-(a)masi

The negative underwent some changes in the transition between OJ and EMJ which were outlined in 3.1.4. 8.3. In the course ofEMJ, the negative adnominal -nu (e.g. kaka-nu 'doesn't write') acquired a reduced variant -n (kakan) which is still used widely in Kansai today. As described above (8.2.1), the negative auxiliary acquired a secondary conjugation, -(a)zar-, which was used especially for combination with tense and mood auxiliaries. 8.4.1

Passives and causatives

During the EMJ period, the OJ passive -(a)ye- and causative -(a)sime- went out of use and were replaced by the alternative OJ passive -(a)re- and the causative -(a)se- which is only incipiently and partially attested in OJ (3.1.4.4). Passive -(a)ye- was lost from early on in EMJ, but is reflected in lexicalized forms such as cNJ mie- 'be visible'(< OJ mi-ye- 'see-PAss') and in fossilized

8.4 Auxiliaries

237

fonns such as arayuro 'all'(< ara-yuro 'exist-PASS~ can exist') (cf. 9.1.7)~ it was also used a little in kanbun-kundoku. Causative -(a)sime- gradually declined through EMJ, but was retained archaically in kanbun-kundoku and in kanji-kana majiribun writings which were heavily influenced by kanbunkundoku. However, it was used as a respect auxiliary through the EMJ period (see 12. 7). An important change in function of the passive is that from EMJ onwards it also came to be used to express respect~ like OJ -(a)sime-, the new causative was also used to support respect auxiliary verbs (see 12.7.1.2). In OJ, passives were only fonned on consonant base and UM verbs, but not on LB, UB, s-irr and k-irr verbs, and that pattern holds also for the few examples of OJ -(a)se- ( cf. 3.1.4.4). However, since EMJ, both causatives and passives are freely fonned on verbs from all verb classes. The shape of the passive and causative auxiliaries which came to be used with LB, LM, UB, s-irr and k-irr verbs was new: -rare- and -sase-, and must have arisen by some sort of analogy. They were also extended to use with UM verbs, such that -rare- and -sase- were used with all vowel base verbs~ the incipient OJ causatives attested on a few UM verbs: Ieise- 'make wear~ dress' and mise- 'make see~ show' were lexicalized with the meanings 'dress' and 'show', and productive EMJ causatives were formed on these verbs as kisase-, misase-. The traditional analysis of the fonnation is that each auxiliary has two basic variants, -sase---(a)se- and -rare---(a)re-, with the long forms attaching to the base of vowel base verbs (ake + sase~ ake +rare) and the short fonns attaching to the a- stem of consonant base verbs (e.g. kaka + se, sina + se~ kaka-re, sina-re), but the analysis usually assumed for NJ is equally possible for EMJ, positing for each auxiliary a unifonn shape, -sase- and -rare-, whose initial consonant is deleted after a base final consonant (e.g. kak + sase => kakase-, sin+ sase => sinase~ kak +rare=> kakare-, sin+ rare => sinare). (2)

Causative -sase-

Passive -rare-

'write' 'die' 'be'

kakasesinasearase-

kakare sin arearare-

'open (tr.)' 'kick' 'arise' 'see' 'come' 'do'

akesasekesaseokisasemisasekosasesesase-

akerarekerareokiraremirarekorareserare-

As a result, the passive and causative auxiliaries became entirely parallel, in distribution (now occurring with all verb classes), and in fonn and distribution

238

8 EMJ granunar Table 8.8 EM/ forms of causative and passive auxiliaries

Base

Causative

Passive

-sase-

-rare-

sasu sasuru sasure saseyo sasezi sasebaya

raru raruru rarure rareyo rarezi rarebaya rarenamu na VERB-rare so

sase sasete sasetutu saseba sasureba sasuredo

rare rarete raretutu rareba rarureba raruredo

Finite

Conclusive Adnominal Exclamatory Imperative Negative conjectural Optative Optative Prohibitive Non-finite

Infinitive Gerund Continuative Conditional Provisional Concessive

of variants (differing only in their consonants: s for causative and r for passive); this is very different from OJ where there was no obvious morphological parallelism between causatives and passives. The EMJ causatives and passives seem to reflect a further morphologization of the derivational suffixes -(a)s'transitive' and -(a)r- 'intransitive'. The causative and passive auxiliaries belong to the LB conjugation; they exhibit the forms shown in Table 8.8, i.e. almost all the forms of lexical verbs. As in OJ, combinations of causative and passive are not found when they are used in causative and passive functions, except for rare examples such as awaserare-, where awase- 'join'< aw- < ap- 'meet' could be taken as a lexicalized verb rather than a productively formed morphological causative. However, from early LMJ onwards they did combine when used in exalting function (see 12. 7).

8.4.2

Aspect; stative and perfective

The main change which took place in the inventory of forms of the aspect and tense auxiliaries compared with OJ (3.1.4.5) was the change of the periphrastic stative -(i)te ar- (cf. 3.1.4. 7.3) to a stative auxiliary, -(i)tar-. In OJ the periphrastic stative sometimes showed phonological fusion, VERB-tear-=> -VERBtar-, and in the course ofEMJ the fused shape was established as an independent

239

8.4 Auxiliaries Table 8.9 Basic paradigm of the EM.! stative auxiliary -(i)tara- stem onbin stem

tarataQ

Infinitive Conclusive Adnominal Exclamatory Imperative

tari tari taru tare tare

stative auxiliary which belongs to the r-irr conjugation and has the basic paradigm in Table 8.9. Morphophonologically, this change is similar to the emergence of the secondary conjugations of the adjectival and regular copulas and the negative auxiliary. It should also be noted that periphrastic stative constructions remained in use through the MJ period, formed by a verb of existence (ar- or one of its synonyms) following the verbal gerund. The establishment of stative -(i)tar- is an important part of a shift which began in EMJ in the system of tense and aspect, in which original aspectual markers, stative -{i)tar- and perfective -{i)te-, acquired tense uses, and eventually in subsequent periods supplanted the original tense markers, -(i)ki and -(i)ker- (
240

8 EMJ granunar

In OJ, the perfective auxiliaries, -te- and -n-, did not combine with the stative -yer- and that pattern continued in EMJ. However, they did combine with the new stative/perfect auxiliary -(i)tar-. Significantly, they attached differently: -ni-tar-, but -tari-te-. This is related to uses of the perfectives not observed in OJ: their uses as perfectives as described above (3.1.4.6) continued in EMJ in what Takeuchi terms 'sequenced narration', but by contrast, in 'embeddings' -(i)te- marks 'recent past tense' whereas -(i}n- marks 'limited control', i.e. the accomplishment of an event 'with (considerable) difficulty or unintentionally' on the part of the agentive subject (Takeuchi 1987: 135). In these functions the distribution of -(i)te- and -(i)n- was markedly different from in OJ. They both (a) combined with -(i)tar-, and (b) each combined with all types of verb, irrespective of the transitivity of the host verb. 8.4.2.1 Progressives OJ had an analytic progressive consisting of infinitive + war- (the opaque lexicalized stative form ofwi- 'sit down'), cf. 3.1.4.7.4. This construction was in EMJ paralleled by several similar constructions, in which a stative form of a verb of vertical motion, such as pus-er- 'be lying down' (cf. pus-lie down'), tat-er- 'be standing' (tat- 'stand up'), or, most frequently, wi-tar- 'be sitting' (wi- 'sit down'), was attached to the infinitive of a verb, e.g. mi witari 'is looking', mi tateri 'id. ', ipi witari 'is saying', katari puseri 'is telling'. These constructions are both interesting in their parallelism with the OJ progressive and as links to later progressive/stative constructions as well as changes among existential verbs ( cf. 12.4).

8.5

Modality

The conjectural auxiliary -(a)m- occupies a central position in the morphological system of modals in OJ and EMJ. In the course of the onbin sound changes it acquired variant forms for the conclusive and adnominal forms, see (3). The vocalic onbin form, -ii, is not used much in the texts until towards the end of EMJ, but it is the form that eventually prevailed and which is used in LMJ and reflected in NJ (cf. 12.1.3). (3)

Conclusive Adnominal Exclamatory Concessive

mu ~ n/ii mu ~ n/ii me medo(mo)

In EMJ, the already rich OJ inventory of grammatical modal expressions was further augmented. The new inflected optative verb form in -(a)baya was mentioned above (8.1.1). Some other new forms incorporate the OJ conjectural

8.5 Modality

241

-(a)m-: (a) The OJ combination of simple past and conjectural had at the EMJ stage contracted to a simple auxiliary, past conjectural -(i)kem-, inflecting like -(a)m-. (b) Anew optative auxiliary -(a)maposi-, e.g. mi-maposi 'want to see' arose from truncation of the OJ construction -(a)maku (no) posi, e.g. mimaku posi or mimaku no posi 'want to see', that is, conjectural -(a)min the nominal form, functioning as a complement to the adjective posi. -(a)maposi (> -(a)mawosi- > -(a)maosi-) belongs to the adjectival shiku conjugation; it is very rare in the first half of EMJ, but used widely later. OJ -(a) maku was also used as the complement of other adjectives; with the adjective u- 'sad, undesirable' it was contracted to -(a)mau- described as a negative optative, but this auxiliary is almost entirely found in the Genji monogatari. (c) From truncation of the construction -(a)mu (to) se- 'coNJ (COMP) do; be about to, intend to' arose the auxiliary -(a)nze-- -(a)iize- 'intentional' which belonged to the s-irr conjugation. As with the conjectural auxiliary, the vocalic onbin form -(a)iize- appears later in the texts than the consonantal -(a)nze-, but was the form which won out and is reflected in LMJ. The intentional overlapped in function with the conjectural. Although it became frequent, it was also fairly short-lived and was used only in the second half of EMJ and in LMJ, but disappeared early in NJ (cf. 12.1.3). Among the extensions, OJ masizi- negative potential was abbreviated to give EMJ mazi- which is usually described as the negative counterpart of be-. A new evidential extension mer- 'seems, appears' came into use in the first half of the period; it is usually thought to originate in a contraction of mi 'see' or me 'eye' and ar- 'be', and it belongs to the r-irr conjugation. The OJ presumptive extension rasi- lost its inflection and became a final particle, rasi, which, however, went out of use before the end of EMJ. In addition to bekasimentioned above (8.2.2), another short-lived formation on be- was herawhich belonged to the adjectival nouns and was used only in the first half ofEMJ. Of the OJ final particles (cf. 3.7.5), kama was lost, but functionally replaced by the new exclamatory particle kana. Also evidential miyu was lost, functionally to some extent replaced by the new extension mer-. A new desiderative particle gana was used only after the combination of perfective and simple past auxiliaries, e.g. mi-te-si gana 'see-PERF-SPST.ADN DESIO; they wanted to see her' (Taketori). Also a new emphatic particle kasi came into use. Modal forms thus include inflected forms, auxiliaries and extensions. They were an important and conspicuous part of the EMJ language. The EMJ inventory is as in (4). No comprehensive analysis of the modal system(s) involved is available, so the forms are simply listed here with their traditional labels, but even in this form ( 4) gives an impression of the amount of morphologically encoded modality in the language.

242

8 EMJ granunar

(4)

Inflected forms -(a)baya optative, -(a)namu optative, -(a)zi negative conjectural, -el-yo imperative. Auxiliaries -(a)m- conjectural, -(a)nze-1-(a)flze- intentional, -(a)maposioptative, -(a)mau- negative optative, -(a)masi subjunctive, -(i)kem- past conjectural, -(i)ker- modal past. Extensions be- necessitive, bekasi- 'id.', mazi- negative necessitive, ram- present conjectural, mer- evidential (sight), nar- evidential (hearsay, sound). Final particles gana desiderative, kana exclamatory, kasi emphatic, mogana optative, na prohibitive, rasi presumptive, yo emphatic.

8.6

Derivatives

A number of derivatives appeared during EMJ which to some extent overlap in meaning, expressing some kind of evidentiality or likelihood and likeness. They include those shown in (5). Most went out of productive use after EMJ, but -gar- and -ge- are still in frequent use today (cf. 12.1.5.3). (5)

-gamasi- 'seem like, resemble' (derives an adjective from nouns or the infinitive of verbs); -gapasi- (> -gawasi-) '-prone, be likely to' (derives an adjective from nouns or the infinitive of verbs); -gar- 'seem, appear, behave like' (derives a verb from adjectives, adjectival nouns, or nouns); gati '-prone, be likely to' (derives an adjectival noun from nouns or the infinitive of verbs); -ge 'seem, appear, behave like' (derives an adjectival noun from adjectives, adjectival nouns or nouns); -mek- 'seem, be like' (derives a verb from adjectives, adjectival nouns or nouns); -yaf1 '-like' (derives an adjectival noun from nouns; SJ loan (:1*), cf. 9.2.3.2).

8. 7

Particles

The case particles ofEMJ are shown in (6), which is very similar to the inventory of case particles in NJ. Of the OJ case particles (see 3.7.1), genitive tu and na ( ~da) were lost, but reflected in lexicalized nouns (ama-tu-sora (heaven +sky) 'sky', OJ ma-na-kwo > EMJ manako (eye+ ?child) 'eyeball'). The rare nominative particle i was lost in the transition between OJ and EMJ, although it continued to be used archaically - and quite frequently - in kanbun-kundoku.

8.7 Particles

243

Ablative kara and allative pe (< OJ pye) became firmly established in this period, and the OJ restrictive particle made came to be used as an allative particle. Of the different variants OJ ablative ywo(ri) ~ yu(ri), only yori was used from HMJ onwards. OJ ni and nite became more specialized with ni being used more for arguments and nite more for adjuncts; from mid to late EMJ nite acquired the variant de which is still in use in contemporary NJ. Note that ni, no, nite (de) and to in addition to their uses as case and conjunctional particles remained forms of the copula, as they do in contemporary NJ. (6)

Accusative Genitive Dative Comitative Instrumentalllocativ e Ablative Ablative Allative Allative

WO > 0 ga, no ni to nite ~de yori kara pe>we>e made

In the course of regular sound change (-p- >-w-and -w- > 0, cf. 7.3), accusative wo > o, allative (OJ pye >) pe > we > e and topic pa > wa. This shows that these particles, and probably particles in general, were phonologically integrated with the host word, for the sound changes affected only 1-p-, -w-1 in non-initial position. 8. 7.1

Genitives

Genitive ga and no continued to mark subjects in subordinate clauses and nominalized main clauses as in OJ (cf. 3. 7.1.1.1). Ga, which in OJ was used only with pronouns and nouns referring to humans, became even further restricted through EMJ, eventually occurring almost exclusively with pronouns, nouns used for 1st or 2nd person reference, and personal proper nouns. Conversely, in EMJ no seems to begin marking subjects in declarative, nonnominalized main clauses. See 12.6.2 about subject marking and the development of a nominative particle. However, as in OJ ga rather than no was used to case mark nominalized clauses whose predicate was in the adnominal form and this is reflected for example in the development of ga as a conjunctional particle (see 8.7.2). Especially from EMJ, both no andga have a function which is often referred to as 'pronominal', as in (7) and (8). There are a few such examples in OJ, but more from EMJ onwards. Ga was used in this function for as long as it was used as a genitive, i.e. into LMJ, and no is still used in this way in NJ,

244

8 EMJ granunar

see (9). This use of genitives is probably simply an instance of genitive noun modification where the modified noun is deleted, so that for example (7)-(9) derive from (lOa-c). (7)

1ma no aruz1 mo saki no mo te now GEN governor ETOP before GEN ETOP hand tori-kapasite take-join.GER 'The current governor and the previous one joined hands' (Tosa)

(8)

ko this

(9)

wa Hanako kono hon this book TOP Hanako 'This book is Hanako's'

no uta pa aru pito no ipaku GEN poem TOP some person GEN say.NOM ~potom o no kuronusi ga nari GEN Kuronushi GEN COP.CONCL Otomo 'This poem is, according to some people, Otomo no Kuronushi's' (Kokin~ remark after 17.899)

no GEN

da COP

(10) a. ima no aruzi mo saki no aruzi mote tori-kapasite b. ko no uta pa ... Opotomo no Kuronusi ga uta nari c. kono hon wa Hanako no hon da.

8. 7.2

Conjunctional particles

As in OJ, conjunctional particles followed the adnominal form of verbs and adjectives. The directional case particles made, yori and kara could be used after nominalized clauses in conjunctional functions 'until', 'from, since', 'after'. In the sense 'because' kara-ni came to be used in the second half ofEMJ. In addition to wo, ni and to, also ga came to be used as a conjunctional particle 'and, as, but', following the adnominal form of a verb. (11) from the Konjaku monogatari-shii is an early example: (11)

oti-iri-keru toki m1 no toki bakari fall-enter-MPST.ADN time hour.of.the.snake GEN time around ga p1 mo yauyaku kure-nu nari-keru COP-MPST.ADN GA day ETOP gradually get.dark-PERF.CONCL 'It was around the hour of the snake (10 a.m.) when he fell in, and/ but (now) it was gradually getting dark' (Konjaku 16.24)

8.8 PronoWls and demonstratives

245

This usage is thought to have developed through reinterpretation of headless nominalized clauses (cf. 12.6.1.3) marked(as subjects) by ga, as clauses linked by ga to the following clause (cf. Kinsui 2007). For example, (12) is open to the two interpretations shown. On interpretation (a) kami ito kiyora-nite nagakarikeru is a headless nominalized clause, of the kind often referred to as an internally headed relative clause (see 12.6.1.3), meaning 'hair, which had been very beautiful and long' and marked by ga as the subject for oti posorite~ but on interpretation (b) it is a coordinate clause connected to the following clause by ga. (12)

a. b.

8. 7.3

kami ito kiyora nite naga-kari-keru ga hair very beautiful COP.GER long-ACOP-MPST GA wake-tori-taru yafi m oti posorite divide-get-STAT.ADN appearance COP.INF fall become.thin.GER 'Her hair, which had been very beautiful and long, had fallen out and become thin as if it had been taken away' 'Her hair had been very beautiful and long, but (now) it had fallen out and become thin as if it had been taken away' (Genji: Makibashira) Tote, nado

A new particle tote appears in this period, which functions as a variant of the particle to in its functions of complementizer and purposive conjunctional particle 'in order to'~ like to it often occurs without a following governing verb of utterance or thinking, meaning 'saying, thinking', and it is for that reason often described as originating in a contraction of to ipite 'saying that' or to omopite 'thinking that', but it is more likely that it arose as a gerund-like variant of to in parallel with ni ~ nite, reflecting that most or all functions as a grammatical particle of to were grammaticalized from its primary copula function. This finds further support in the new EMJ forms nado ~ nadote 'why' which are etymologized as < nani-to ~ nani-tote 'being what?'.

8.8

Pronouns and demonstratives

The OJ simple system of personal pronouns was already in OJ supplemented by a number of terms of address (cf. 3.8.2), and it was lost in the course of EMJ. While the OJ 1st person pronoun ware/wa-ga continued in EMJ, the OJ 2nd person nare/na- went out of use, and around the middle of EMJ there is no longer a system of personal pronouns as such, but an inventory of terms of address and of self-reference. The development has aptly been described as one of 'de-pronominalization' (Vovin 2003: 95), accompanied by the extensive

246

8 EMJ granunar

Table 8.10 EMJ demonstrative and interrogative forms

Short Long Location Direction Location/direction Manner Time Quantity/degree

Proximal (speaker)

Mesial (hearer)

lw lwre lwlw lwnata lwti kaku(te), kau (ima)

so sore solw sonata soli sa(te), sika

Distal

Interrogative

kala karelare kasilw kanatalanata

idure iduku- idulw - idolw - dolw

(ka-yaiJ)

-

idutilidura ika, ikaga, ikani, ikade itu ikura

use oftenns referring to the speaker also as reflexives. The fonns in (13) were used for 1st, 2nd and 3rd person personal and reflexive reference. Several of the forms involve OJ onore/ono 'reflexive' and the nounmi 'body'(< OJ mwi). The personal interrogative pronoun tare 'who' and the interrogative noun nani 'what' remained in use. (13)

1st person: mara, mi, onore-ga-mi/ono-ga-mi 2nd person: nandi (< OJ namuti), kimi 3rd person: watakusi (originally 'private') 1st and 3rd person: ware/wa-ga, wa-ga-mi, midukara 1st, 2nd and 3rd person: onorelono-ga, onodukara 2nd and 3rd person: omape (> omawe > omae), gozen 2

8. 8.1

Demonstratives and interrogatives

The OJ two-term demonstrative system of speaker versus non-speaker forms (3.8.3) changed into the three-tenn system familiar from NJ. Main forms, including the corresponding interrogative forms, are shown in Table 8.10. It is thought that the distal terms originate as a subdivision of the OJ proximal, visual domain into a close (proximal) and distant (distal) distinction, and that the OJ non-speaker based terms subsequently were reinterpreted as mesial and hearer based. Whereas the non-speaker based forms in so- in the OJ system were mainly used anaphorically, the system from EMJ became clearly deictically focused on speaker, hearer and distal. However, the mesial fonns in so- continued to be used extensively in anaphoric function, and the new distal ' Omape
8.9 Kakari-musubi

247

was also used (semi-)anaphorically in the sense 'that well-known', often in literary or scriptural references. Anaphoric usage is also found with the temporal deictic ima 'now' to mean 'then; at the time'. The morphological system is overall characterized by further regularization compared to OJ, but there was yet not the full parallelism within the demonstrative system, nor between the demonstrative and interrogative systems, which characterize NJ, although for example the new locational interrogatives iduko, idoko and doko clearly show some levelling in form. Konata, sonata, kanata are thought to have arisen from reductions of ko-no kata 'this side', etc.; the distal forms in a- are somewhat less frequent than those in ka-. Both kanata and anata were used for 3rd person reference in EMJ and LMJ; anata only started being used for 2nd person reference towards the end of the eighteenth century, i.e. well into NJ.

8.8.2

Short and long forms

In OJ pronouns and demonstratives had a fairly productive system of short and long forms, e.g. wa/ware and kolkore (3.8.1). In EMJ the short forms became restricted to use only with a genitive particle (wa-ga or ko-no).

8.9

Kakari-musubi

Kakari-musubi ('hanging-tying') is the name in Japanese grammar for a construction in which some constituent is marked by one of the 'kakari particles' (a) ka, ya, solzo, namolnamu or (b) koso, and the sentence predicate it relates to is in the (a) adnominal or (b) exclamatory form, rather than in the conclusive form generally used to conclude sentences. This construction is found in both OJ and EMJ, where it is a conspicuous and prominent feature of the language. It disappeared from the language in the course of LMJ and NJ as part of a complex of syntactic changes (12.6.1.2), but remained a normative feature of classical written Japanese. The construction remains in Ryukyuan and in a few mainland dialects to a more limited extent than in OJ and EMJ. Kakari-musubi has recently attracted a great deal of attention as part of the growing interest in the syntax of OJ and EMJ, and also in the context of reconstructing prehistoric Japanese syntax and morphology. Specifically, within mainly or purely formal syntactic analyses, kakari-musubi is sometimes said to involve movement, which has further been taken to support the analysis of OJ as having overt wh movement (cf. Watanabe 2002), but this is primarily of theoretical interest to formal linguistics. Traditionally, kakari-musubi is thought of as an automatic agreement rule by which the presence of one element (a kakari particle) triggers or requires

248

8 EMJ granunar

the presence of another (a predicate (musubi) in the adnominal or exclamatory form), and the function of the kakari particles is said to be 'emphatic' (zo, namu, koso) or 'interrogative' (ka, ya). (14) gives some examples from OJ; see (8.9.2) about the differences between the kakari particles. (14) a.

no wa ga kwopuru ywo kimi so kizo my.lord so last. night GEN night I GEN love.ADN 1me m mi-ye-turu dream DAT see-PASS-PERF.ADN 'It was you, my beloved lord, that I saw last night in a dream' (MYS 2:150)

b.

sikama-gapa taye-mu p1 ni koso a ga Shikama-river cease-CONJ.ADN day DAT KOSO I GEN kwopwi yama-me love stop-CONJ.EXCL 'It is on the day when the Shikama River ceases its flow that my love for you will end' (MYS 15.3605)

c.

kagurwoki kami ni tuyu so oki-ni-kyeru black hair DAT dew SO fall-PERF-MPST.ADN 'It is dew that has fallen on my black hair' (MYS 15.3649)

d. tukwi wo sirwotape no kumo ka moon ACC white.cloth COP.ADN cloud KA kakus-eru ama-tu-kwiri kama hide-STAT.ADN sky-GEN-mist KAMO 'Is it white clouds that are hiding the moon? Or is it the mist in the sky?' (MYS 7.1079) More recent studies posit and emphasize fundamental functional and syntactic differences between sentences with and without kakari-musubi, viewing kakari-musubi as a focus construction, often with much the same semantic effect as it-clefts in English, as reflected in the translations in (14) above, but differing from cleft constructions by being mono-clausal. Thus, for example (14a) has the following focus-presupposition structure, with the focused constituent followed by the presupposition. This is structurally and semantically a quite different sentence from its non-focused counterpart(s) without kakarimusubi, which could be for example (16a) or (b): (15)

Focus wa ga kwopuru kimi so 'It was you, my beloved lord

Presupposition kizo no ywo ime ni miyeturu that I saw last night in a dream'

8.9 Kakari-musubi

249

(16) a. kizo no ywo ime ni wa ga kwopuru kimi miyetu b. kizo no ywo wa ga kwopuru kimi ime ni miyetu 'Last night I saw you, my beloved lord, in a dream' The traditional term, kakari-musubi, which dates from the nineteenth century, makes reference to two sentence elements: kakari means 'hanging (down, out); relation, attachment, connection' andmusubi refers to concluding ('tying up, together') a sentence; kakari-musubi may be thought of as that which is being 'hung out' (kakari) and that whereby it or the sentence is being 'tied together' (musubi). This conception of kakari-musubi is not dissimilar from that of the 'theme-rheme' relation (cf. Greek thema 'that which is set up', rhema 'that which is said'); note in particular that kakari and thema are very similar in meaning. In fact, Moto'ori Norinaga's original classification (in his Kotoba no tama no o, 1777/85) includes wa (contrastive) topic and mo inclusive topic, and zero (in zero marked topics) among the forms that take part in kakari-musubi, and today, the 'kakari particles' of Classical Japanese in Japanese school grammar comprise wa, mo, ka, ya, zo, namu, koso. For zero, wa (< pa), and mo the musubi (predicate) takes the conclusive form, whereas it takes the adnominal or exclamatory form with the other kakari particles. 'Kakari-musubi' was first used in this wider sense (in Togashi Hirokage's Kotoba no tama-hashi, 1826); current use of the term to refer only to the relationship between ka, ya, zo, namu, koso and the form of the predicate to which they relate is more recent. Today 'theme-rheme' has in most branches of linguistics been replaced by 'topic-comment', but there are differences in the original meaning of the terms, and topic-comment could be viewed as a subclass of theme-rheme. Pursuing the older conception of kakari-musubi, it is possible, without stretching the Japanese or European terms too much, to equate kakari-musubi and theme-rheme, as in (17), where topic and focus constructions are viewed as different subtypes of kakarimusubi (theme-rheme): (17)

Kakari (theme)

Musubi (rheme)

Topic

Comment conclusive

0, wa, mo

Focus ka, ya, zo, namu koso

Presupposition adnominal exclamatory

This understanding of a relation between a kakari and a musubi constituent is, however, probably too broadly conceived, for the two types of kakari, topic and focus, are not mutually exclusive, and from a functional point of view

250

8 EMJ granunar

they belong to different levels of representation: topic-comment belongs to the utterance level, and focus-presupposition to the level of proposition. Sometimes the referent of a focused constituent will correspond to a kind of topic at utterance level (though not a backgrounded, 'old information' -type topic), as in (18) where the focused constituent is also marked as a contrastive topic. (18)

ima koso pa wa-dori ni ara-me now KOSO TOP my-bird COP.INF exist-CONJ.EXCL '(It is) now (that) I am/will be my own' (as opposed to later when I will be yours) (KK3)

Often, however, topic and focus are separately represented, with structures such as (19). For example, (14) above may in fact be interpreted in that way, see (20). (19)

Topic

Comment Focus

(20)

sikama-gapa [0] 'The Shikama river,

tayemu pi ni koso a ga kwopwi yamame it is on the day when that my love for you it ceases its flow will end'

kagurwokikamini[0] tuyu so 'On my black hair, it is dew tukwi WO (0] 'The moon,

Presupposition

okinikyeru that has fallen on it'

sirwotape no kumo ka kakuseru is it white clouds that are hiding it?'

A focus construction establishes a copular, predicative relation between focus and presupposition, like that between subject and nominal predicate, see (21), much more so than does for example the topic-comment relation, which is also a 'connecting' relation, but not a copular one. This shows one significant difference between topic and focus constructions. Like focus-presupposition, the subject-(nominal)predicate relation belongs to the level of proposition. (21)

Focus

IS

presupposition

[wa ga kwopuru kimi so] you my beloved lord

IS

[kizo no ywo ime ni miyeturu] (what) I saw last night in a dream

[tuyu so] dew

IS

[okinikyeru] (what) has fallen

251

8.9 Kakari-musubi

8. 9.1

[tayemu pi ni koso] on the day it ceases

IS

[a-ga kwopwi yamame] (when) my love will end

[sirwotape no kumo ka] white clouds

IS

[kakuseru] (what) are hiding (it)

Uses of kakari-musubi

Kakari-musubi is structurally a focus construction, but in many cases, especially in poetry, the construction seems mainly to contribute exclamative or interrogative force to the utterance, or nuances of doubt or various kinds of emphasis; although we illustrated the basic structure by translating sentences with kakari-musubi by it-clefts above, it would in many cases be contrived or misleading to do so. See for example (22), a poem from the Ise monogatari, which has no fewer than three kakari particles. (The last two stanzas (yuki . .. pa) are a right-dislocated complement of omopiki.) Here the two uses of kakari-musubi in yume ka to zo omopu express 'emphatic doubt' ('what I wonder is, is it a dream!'), emphasizing the feeling of unreal-ness which the loyal servant experiences upon seeing his former master in humble, out-of-the way dwellings. (22)

wasurete yume ka to pa zo omopu forget.GER TOP dream KA COMP zo think.ADN ya omopi-ki think-SPST.CONCL YA yuki pumi-wakete kimi wo m1-mu snow stamp-divide. GER my.lord ACC see-CONJ.CONCL to pa COMP TOP 'Forgetting [that you have retreated from the world], this seems unreal! (I wonder, is this a dream?)! That I would have to stamp my way through the snow to see my lord- that I never thought [did I think that? (No!)]' (Ise 83)

8.9.2

The individual kakari particles

There are significant functional/semantic differences between the individual kakari particles and the sentences they constitute. Semantic tags (from Quinn 1987) and some salient distributional facts about the kakari particles are summarized in (23), which shows whether they occur in questions, are used with wh- words, and whether they can be used in sentence final (SF) position, in which case they focus the entire proposition.

252

8 EMJ granunar

(23) ka 'doubted identity' ya 'confinnation soliciting' zo 'identifying' namu 'confinnative' koso 'unique identification'

Q

Wh-

SF

+ +

±

±

±

+ + +

Ka, ya are only used in questions, whereas zo may be used in questions; ka, zo, but not ya, may be used with wh- words. Ka, ya, zo are also used sentence finally (and are in this use usually classified as 'final particles' in Japanese school grammar); when sentence final ka, ya, zo follow a nominal predicate there is usually no overt copula between the noun and the particle. Namu, koso are not used in questions, with wh- words or sentence finally. In Quinn's terminology, ka, zo, koso are said to involve identification, that is, establishing (zo, koso) or establishing and questioning (ka) an identity relation between the focused constituent and the presupposition, whereas ya, namu either seek (ya) or provide (namu) confirmation of the relation. Semantically and in tenns of distribution ka/koso seem to be paired as opposites, as are ya/namu. Zo is used most widely, overlapping with both ka and koso. Ka ('doubted identity') is used only in questions, in order to focus a constituent, often a wh- word (24) ('who, what, when, how, ... is it that ... ?'), but it can also be a lexical word singled out as the focus of a yes/no question, (25) ('is it X that ... ?'). When ka is used sentence finally, it is the predicate which is focused, (26).

(24)

kakaru miti pa ikade ka imasuru KA be.RESP.ADN be.like.this.ADN path TOP why 'Why is it that you are on a path such as this?' (Ise 9)

(25)

kurenawi no mo no suswo nurete ayu ka cnmson GEN robe GEN hem get.wet.GER sweetfish KA turu ramu catch.CONCL PCONJ.ADN 'the hems of their crimson robes being wet, is it sweetfish that they are catching?' (MYS 5.861)

(26)

pa pana ka ara-nu ka siragiku white.chrysanthemum TOP flower KA exist-NEG.ADN KA nam1 no yosuru ka wave GEN approach.ADN KA 'The white chrysanthemums, are they flowers, or not? Are they breaking waves?' (Kokin 5.272)

8.9 Kakari-musubi

253

Ya ('confinnation soliciting') is used in yes/no questions which seek the hearer's confinnation that an already identified state of affairs (the presupposition) is to be linked with the constituent inside the sentence marked by ya ('speaking of X, is it true that ... ?'), see (27)~ as here, ya is often used in disjunctions ('is it A, or is it B ?). Ya can also be used in wh- questions to focus a non-wh- element, e.g. (28). Ya is often used in rhetorical questions, with exclamative force, (29). At the end of a sentence, ya solicits confirmation of the sentence in its entirety, that is, marks a neutral yes/no question. This is different from other uses of the kakari particles, for it involves no discernible focus, and this is reflected in the fact that the predicate preceding ya is in the conclusive fonn, not the adnominal, see (22) above. Sentence final ya is also often used in exclamative rhetorical questions, in OJ also after the exclamatory fonn, (30). (27)

kimi ya ko-si ware ya yuki-kemu you YA come-SPST.ADN I YA go-PSTCONJ.ADN 'Was it you that came, or me who went, I wonder' (Ise 69)

(28)

koko ya idoko here YA where 'Where is this~ where are we' (Tosa)

(29)

ya ara-nu paru ya mukasi no paru tuki moon YA exist-NEG.ADN spnng YA old.times COP.ADN spnng nara-nu COP-NEG.ADN 'Isn't the moon the same~ isn't the spring that of old?' (Ise 4)

(30)

mo mtpune no yama ni tayuru sirakumo white-cloud ETOP Mifune COP.ADN mountain DAT vanish.ADN p1 ara-me ya day exist-CONJ.EXCL YA 'The white clouds will never disappear from Mt. Mifune! (Will there be a day when the white clouds disappear from Mt. Mifune? (No!))' (MYS 3.243)

Zo (so) ('identifying') identifies a constituent for focus, as new information on the background of the presupposition which follows ('it is X that ... '), e.g. (31). Like ka, zo can focus an entire sentence, e.g. (32). In OJ, the shape so is more common, the variant zo less so, but it is thought that zo in EMJ had replaced so, and the particle is commonly referred to as zo. We follow that practice here, but as most EMJ texts did not distinguish

254

8 EMJ granunar

between tenues and mediae in writing it is difficult to agree that the replacement of so by zo was as unquestioningly complete in EMJ as is usually assumed~ note also for example that the Ruiju-my8gi-sh8 (6.2.3) have unambiguous notations of the particle as so. (31)

kyaii pe kaperu ni wonnago no capital ALL return when girl GEN na-ki nomi zo kanasibi-kopuru not.exist-ACOP.ADN only zo grieve-regret.ADN 'Returning to the capital, it is only that my daughter is no more that I grieve and regret' (Tosa)

(32)

wa ga korom o sur-eru m pa I GEN clothes colour-STAT.ADN COP.INF TOP ara-zu exist-NEG.CONCL takamatu no nwobye yuki-sikaba pagw1 no Takamatsu GEN field go-SPST.PROV lush clover GEN sur-eru so colour-STAT.ADN SO 'It is not that I coloured my robe. It is that the bushclover rubbed off on it, when I was walking in the fields in Takam atsu.' (MYS 10.2101)

Namu (namo; nan) ('confirmative') is used to solicit agreement from the hearer and is typical of very 'engaged registers of speech' (Quinn 1987: 741) and is often used in explanations ('it is X, you see, that ... '), e.g. (33), in several respects resembling NJ ne(e) 'you see'. It is therefore frequent in the literary prose, especially in conversations and in personal narrative styles (and narrator comments), but was not used in poetry. Thus, in OJ, namo (which was the OJ shape) is not frequent and is restricted to Senmyo (and a single example in MYS (12.2877)). In EMJ, OJ namo had become namu which in some cases was further reduced to nan, e.g. (34). (33)

past wo yatu watas-eru m yorite namu bridge ACC eight lay-STAT.ADN DAT depend.GER NAMU yatupasi to ipi- keru Yatsuhashi COMP call-MPST.ADN 'It is because there are eight bridges, you see, that it is called "Yatsuhashi"' (Ise 9)

8.9 Kakari-musubi

(34)

255

so no pito katati yon pa kokoro nan that GEN person appearance ABL TOP heart NAMU masari-tari-keru excel-STAT-MPST.ADN 'That person, more than her appearance, it was her heart, you see, that excelled' (Ise 2)

Koso ('unique identification') is the only kakari particle whose predicate is in the exclamatory and not the adnominal. This is the case for all types of predicates from HMJ onwards, but in OJ, adjectival predicates of koso focused constituents were in the adnominal, e.g. (36). As opposed to zo, which also 'identifies', koso does so 'uniquely' ('it is X (and only X) that ... '), corresponding to 'exactly' or 'just', see (35), and further (18), (20) above. (35)

ko no wonna wo koso e-rne wotoko pa man TOP this GEN woman ACC KOSO get-CONJ.EXCL to omopu COMP think..CONCL 'The man thought that it was this woman (and her alone) that he wanted' (Ise 23)

(36)

wo piro-mi kusa koso sige-ki nwo moor ACC wide.INF grass KOSO abundant-ACOP.ADN 'with the moor being wide, it is the grass which is abundant' (MYS 17.4011)

8.9.3

The musubi predicate

First of all it should be noted that a distinct form of the kakari-musubi predicate is not expressed in a number of cases: (a) If the final inflected element in the predicate belongs to the QD conjugation (including the frequent conjectural auxiliary -(a)m-), which has no (segmental) expression of the distinction between conclusive and adnominal (it is sometimes speculated that the adnominal was prosodically distinct from the conclusive form also outside its adnominal use, but that is not known; see 7.4.4.3). (b) If the sentence final predicate is incomplete, either abbreviated or left out. (c) If the kakari-musubi construction is within a non-finite clause, as conclusive, adnominal and exclamatory are finite verb forms. Such cases make up a non-neglible proportion of kakarimusubi constructions, suggesting that the distinctive form of the musubi predicate is of less importance than it is often afforded. In most accounts of kakari-musubi within a focus interpretation of the construction, whether they are formally or functionally oriented, the use of the

256

8 EMJ granunar

adnominal or exclamatory form in the corresponding predicates is said to mark the scope of the focus constituent, that is, to show which is the predicate in the presupposition. This is in principle no different from the automatic rule of the traditional agreement description. A notable exception is the work of Quinn (1987) which divorces the two parts of the kakari-musubi construction, the kakari particle and the form of the musubi predicate, and gives each of them a functional, synchronic explanation. Quinn relates the use of the adnominal for the presupposed predicate to its overall function as a nominalized, and therefore 'referential' and 'grounded', verb form, suitable for a presupposed constituent. 3 This explanation of the use of the adnominal form in kakarimusubi constructions is attractive, but does not extend to the use of the exclamatory. The main uses of the exclamatory as a word form, distinct from its uses as a combinatory stem in word formation (cf. 3.4.4.3), were in kakarimusubi with koso, and in exclamations (mostly couched as rhetorical questions and followed by ya, yamo, yapa, ka, kamo), so it is not possible in the same way as with the adnom inal to generalize from other uses of the exclamatory. Most scholarship agrees, though, that there are close diachronic and functional links between the adnominal and the exclamatory and that both exhibit nominal(ized) characteristics. Finally, in syntactic terms, the use of a nominalized verb form in the predicate of the presupposition ties in very well with the copular nature of the kakari-musubi relation, see (31) above. 8. 9.4

Questions and exclamatives

Two other uses of the adnominal form should be mentioned in this context, namely in questions and in exclamatives. As outlined above, yes/no questions are constituted by the presence of ya or ka (with sentence final ya following the conclusive form making up a neutral, unfocused yes/no question). Whquestions, on the other hand, do not require the presence of a question particle, but can be constituted simply by the presence of a wh- word, see (37). Also in such cases the predicate takes the adnominal form, despite the absence of a kakari particle. Wh- words can be considered focus constituents in their own right and that may account for the use of the adnominal form in such sentences, but it has also been suggested that the adnominal form more generally is used, as a marked form, in marked, non-declarative utterance types, including questions, but also in what are interpreted as exclamatives: (38) is one such example where the use of the adnominal form in the predicate is said to make the sentence exclamative. A potential problem with this kind of interpretation 3

Similarly, but in a diachronic perspective, Wrona (2008) proposes that the kakari-musubi construction originated in the function of the adnominal form of marking non-declarative utterance types (interrogative or exclarnative) and that the particles originally were secondary scope markers.

8.9 Kakari-musubi

257

is the absence of any other evidence than the adnominal form that the sentence is in fact intended to be exclamative. There are not many examples of sentence final adnominal form not correlating with a kakari particle in OJ, but numbers increase through EMJ, e.g. (39). (37)

ide ika m kokodaku kwopuru oh how COP.INF this.much love.ADN 'Why do I love so much?' (MYS 12.2889)

(38)

wa ga koromode no toki mo puru I GEN sleeve GEN dry.ADN time FOC na-ki not. exist-ACOP. ADN 'there is no(t even) time for my sleeves to dry!' (MYS 10.1994)

(39)

nigasi-turu suzume no ko o inuki ga sparrow GEN child ACC Inuki GEN let.go-PERF.ADN 'Inuki let my baby sparrows go!' (Genji: Waka-murasaki)

REFERENCES General: lwai 1970, Tsukishima 1969, 1987, Vovin 2003, Yamada 1913/1952. Aspect: Kinsui 2006, Sandness 1999, Takeuchi 1987. Kakari-musubi: Olmo 1993, Quinn 1987: 673--825, 934-49; Watanabe 2002, Whitman 1997, Wrona 2008: 195ff.

9

The sinification of Japanese

Just as early Japanese society received massive intellectual, cultural, political and social influence from Tang China, so the Japanese language was heavily influenced by Chinese, particularly through the MJ period. There is no doubt that prior to this, both in Nara and pre-Nara Japan, some intellectuals and clergy, as well as traders and fishermen, had some facility in varieties of spoken Chinese. As was set out in 4.2.2 above, a number of early loanwords into Japanese from Chinese, possibly mediated through the Korean peninsula, may be identified, and it is likely that there are more which we are not able to identify. However, the pervasive influence on Japanese from Chinese in the OJ and MJ periods which took place through the medium of text was of an altogether different order, affecting both usage and especially vocabulary to an extent which merits the designation sinification. This took place through two related, complementary modes of interacting with Classical Chinese text (kanbun ~Jt), generally thought of as two ways of 'reading' the texts: kanbun-kundoku, the rendition of Chinese text in Japanese, which affected grammar and usage (see 9.1) and (kanbun-)ondoku, the vocalization of Chinese text as such, which paved the way for the intake of a large number of loanwords from Chinese (9.2). Both of these 'reading' practices have a long history in Japan, predating the Nara period and continuing into the present. It is convenient to treat them as one here, for it is from the Heian period we find the earliest direct evidence for the language of kanbun-kundoku and the beginnings of a large-scale adoption of SJ loanwords.

9.1

Kanbun-kundoku

Kanbun-kundoku (~Jt~JIIW'f) is the interpretation, explication or translation in or into Japanese of Classical Chinese text. An important characteristic of kanbun-kundoku is the notion that it involves verbalizing the original Chinese text in Japanese, and it is popularly thought of as 'reading' Chinese text 'in Japanese' or 'with Japanese grammar'. The practice of kanbun-kundoku, understood as 'the "reading" of Chinese in a local vernacular language', is not restricted to Japan, but is a common feature of civilizations within the Sinitic cultural sphere, attested and described in the sixth and seventh centuries from places as far-flung as Japan, the Korean peninsula, Vietnam and Gao Chang (1'1\'J ~, the site of an important oasis city on the Silk Road, in what is now the 258

9.1 Kanbun-kundoku

259

Chinese province of Xinjiang). It is safe to assume that kanbun-kundoku in this wider sense predates these early attestations, and it may be thought that the advent of Chinese text in Japan from the Korean peninsula early in the fifth century was accompanied by this practice. In Korean, rendition of Chinese text in Korean is said to date back at least to the fifth century. It is likely that the strong logographic element in Chinese writing favoured the development and spread through EastAsia of kanbun-kundoku (-like practices), together with the notion that it consists in 'reading' Chinese in another language. Although kanbun-kundoku certainly is a kind of 'translation', we use the broader term 'rendition' in the following in order to capture all of what the practice involves. A basic feature of kanbun-kundoku is the translation of words and phrases in the Chinese text into Japanese. Chinese and Japanese are grammatically quite different: Chinese has no inflectional morphology, expresses grammatical relations by word order and has a large inventory of grammaticalized preposed verbs and adverbs (expressing amongst others categories such as negation and mood), as opposed to the fairly rich verbal and adjectival inflection, specification of syntactic roles by grammatical particles and free word order (except for verb-finality) of Japanese. Thus, kanbun-kundoku involves finding suitable translation equivalents in Japanese for content and function words in the Chinese texts. However, in addition kanbun-kundoku involves a number of processes in order to render ('read') the Chinese text in Japanese: transposition (change of word order) and interpolation or specification (of inflectional morphemes or grammatical particles). 9.1.1

Kunten

Kunten (~}iL~ 'reading marks, glosses') is a cover-term for a variety of annotations added to Chinese text in order to aid these processes of its rendition in Japanese. The earliest extant kunten, from the late eighth century, are punctuation marks, showing phrasing and division of a text, and marks showing how to change the word order when rendering text in Japanese. The latter are collectively known as kaeriten 'reversal marks'; through time these have included numbers and other means of showing sequence. Especially from the Heian period onwards, more types of kunten are found which may roughly be divided into two classes: kana glosses and okoto-ten. Kana glosses are man 'yogana or kana written next to a character, indicating its 'reading'. This could be a SJ word, in which case the gloss only had information about the pronunciation; or it could be the sound shape of a Japanese word used to render the Chinese word in Japanese, in effect constituting a translation or glossing in Japanese of the Chinese word. The development of katakana is closely linked to the practice of glossing, and katakana is traditionally viewed as originating as a subtype of kunten.

260

9 The sinification of Japanese mu

nie

e

eo ekoto

ka/ga e

e no

eto tee

e

ewa

su

Figure 9.1 Example of kunten system

Okoto-ten (thus named after two frequently noted grammatical forms, the particle wo > o and the nominalizer/complementizer koto) or tenioha (named after the gerund formant -te and the particles ni, o and pa > wa (today written using the kana for ha)) are two common terms for diacritic marks which indicate grammatical morphemes. Graphically, okoto-ten/tenioha are lines, dots, circles, hooks or marks of other shapes which were placed next to or on kanji. Okoto-tenltenioha are generally thought of as shorthand for grammatical particles or words, auxiliaries or inflectional endings; both the shapes and positions of marks are significant. Figure 9.1 is an example of part of a system which uses single dot marks for -te, ni, o and wa in the four corners of the space around a kanji, as well as marks for o and koto, exemplifying the frequent markings which gave rise to the two names for this kind of diacritic. It also includes marks for the particles ka, ga, no and to, as well as for su and mu. Through the first half of the He ian period, a wealth of different kunten traditions developed, with individual scholars, sects or temples developing their own systems which grew increasingly complex and often secret, or at least exclusive. Tsukishima (1986) provides a large number of charts illustrating different systems, many of which are very elaborate. Within each school of reading, normative annotations and readings of individual texts became established, and from around the mid Heian period text annotations became the object offaithful, dogmatic tradition. From the Kamakura period, the introduction of neo-Confucianism was accompanied by new text interpretations, leading to some innovation and change in annotations of some Confucian texts. Today a simple system of kunten, confined largely to kaeriten, is taught in Japanese schools as part of the kanbun curriculum.

261 'Ihtfollowingis antxm~plt of ~emndkaWlot·~~. (l)is a Chinut FdhWw.:f~·u((.!~.::t~.Japmut HoAA! 1ftlongu ''Iht strd.t.ruts and phrases in 1ht Lob.ls rutn '), m t~lkWM of t\t Lo1Us Sutra by Zhi·yi (1#!~. Jt;~Q\tst Ch~i) from tht second lu.I of tht sirtl ctmD:y,vnil1m down by ont of his s tudtnts. Zhi·yiwas tht foundtr of Thrd.ti EUddhis:m., tht U\(trtor of Japmut nndt.i Buddhism, md tht PdlwfJ whuU i> mim.portud. to~ in TtndaiBuddhis:m. (2) is t\t ttxtucnot.tttd wflhklmten, adapu d from Nishizd
tt xt,a short. passage fromtht

(1)

1'1;'SA~~.l:.~ ~~£f1m~

;;;~m~1l:r..1:#. t~• ~;~; (2)

£J

n.

~l!t~ A\,

'

.

iii . l'.llr

pg ~~

•• ~

AJ ..~3

~·

-B? - l:Z

!!f." ~!IIU

T

'

-~ ,

>!\~

~

'!!<J ·~

til, ~w-i

J:t

::;~<

262

9 The sinification of Japanese

In (2) we first of all see a great deal of kana glosses, both giving infonnation about 'readings' of individual kanji and supplying grammatical infonnation not represented in the Chinese text, such as particles and verbal auxiliaries and inflectional endings. The kana glosses do not distinguish between tenues and mediae, for example using :JJ both for /ka/ in kaburo and for /gal in genitive ga, and inspection of the photographic reproduction of the original kunten text further shows that a num her of variant kana shapes were used in the glosses. There are also some punctuation marks (a 'comma' after J/Ei, 1~, #~and~; a 'full stop' after~ and :Jit) and kaeriten in the fonn of the numbers 1 or 2 to the left of :flt 1::., W\, ~, Itt, JJ, ~and :Jit, indicating that the kanji (complex) marked '2' should be rendered after that marked '1 '. Finally, there are two instances of okoto-ten, a single dot by the top left corner of a~ and ~, standing for ni. (3) is the yomi-kudasi (lit. 'reading-down'), the Japanese text represented by the kunten. We have noted in CAPITALS readings not indicated in the kunten text, i.e., readings which we hypothesize in our interpretation. Note that #~ appears earlier in the text where it is glossed as riu; 1~ is glossed simply by ke, which suffices to show that it is meant to be glossed potoke and not for example putu. Everything in lower case in (3) is directly represented in the kunten text. We use boldface for glosses giving grammatical infonnation not represented in the Chinese text, and bold italics for such infonnation given by okoto-ten, and we underline words which have been transposed, generally by a move to the right as instructed by numbers, but note also that the order of W\ and 1::. has been reversed without any overt instruction. Knowing and following the conventions used in this tradition of kanbun-kundoku allows a reader to render the Chinese text into Japanese. (3)

1PJ 3tt A ;fJ; 1::. 1Jt JiEJ nazo kaburo naru PITO no WA ga uwe yon suguru, why bald COP.ADN person GEN I GEN above ABL pass.ADN [The Dragon said:] 'Why is a bald man passing over me? ~

NOTI after

no COP.ADN

a~

1~

TOKI ni time DAT

POTOke, Buddha

~

1::.

AME ni heavenDAT

NOBOra-mu to possu ascend-CONJ.CONCL COMP want.CONCL

'Later, the Buddha wanted to ascend to heaven.' ~

#~

m~

KO no this GEN

RIU, dragon

kokuUN-anbu wo black.cloud-dark.mist ACC

r~Jjf

Itt paite breath.GER

263

9.1 Kanbun-kundoku

=.")t

~

SANGWAUwo three.lights (the sun, moon, and stars)

kakusi-kakusu hide-hide.CONCL

ACC

~

'The dragon hid the sun, moon, and stars by breathing black clouds and dark mist.' Several questions arise from kunten texts such as (2). Should they primarily be considered 'annotated' Chinese text with more or less abstract instructions about their interpretation and rendition in Japanese, or are they rather orthographically severely underspecified Japanese text, superimposed on the still present Chinese source text? Is the use of kunten 'annotation' or is it 'writing'? Kunten texts present a multi-layered textuality of great complexity with nontrivial difficulties of interpretation. In China itself, annotating text for interpretation or pronunciation is well established and has a long history. For example, the shoten ('tone marks') mentioned earlier (6.1.2.2) are similar to kunten and are part of the same overall phenomenon. Until recently it was thought that kunten, including katakana, were independent developments in Japan, if perhaps generally inspired by for example shoten; similar types of materials exist in Korea, but are somewhat later. In Korea, marks to annotate text are called kugyol; they are very similar to Japanese kunten, but the earliest have been thought to date from the ninth century. However, especially with the continuing discovery in both Japan and Korea of increasing amounts of kunten materials which are annotated not in ink, but by stylus(~* kakuhitsu) which leaves indentations or scratchings on the paper, but no colour, it is gradually becoming clear that techniques for annotation were used both on the Korean peninsula and in Japan at an earlier time than was previously thought, with the oldest such stylus materials in Korea dating from the late seventh century. It now in fact seems overwhelmingly likely that kunten techniques, too, like Chinese writing and text and kanbun-kundoku, were transmitted from the Korean peninsula to Japan. For example, the earliest Japanese materials are far more similar to the Korean materials than are later Japanese materials. Both kanbun-kundoku and kunten and their histories must be viewed in a pan-East Asian perspective, where, in particular, the spread of Buddhism and Buddhist canonical texts in Chinese translation and commentaries written in Chinese played an important role.

9.1.2

Kanbun-kundoku and writing in Japanese

A close relation holds between kanbun-kundoku and the development of writing in Japanese. In the course of kanbun-kundoku, fixed, habitual

264

9 The sinification of Japanese

renditions of individual kanji arose, resulting in conventional associations of many kanji with specific OJ words; or in other words, the establishment of conventional 'kun-readings' of kanji. Once this association of decoding (reading) was established, the next step of reversing the relation to one of encoding (writing) was not a big one. For example, habitually using Japanese rna- - me 'eye' to translate into, that is read in, Japanese the Chinese word written by § established a representational relation between § and rna- me 'eye': (4)

§=>rna-- me 'eye'

This could now be reversed to have the word rna-- me 'eye' represented, that is, written by, §, see (5), making possible logographic representation of Japanese. This is the origin of logographic writing of Japanese. (5)

rna-- me 'eye'=> §

Furthermore, by extension, once the encoding relation between rna- -me 'eye' and §was established, §could be used as a phonogram (kungana, cf. 1.1.2.5) to write the syllables /rna, me/: (6)

/rna, me/=>§

Both the logographic and phonographic use of § shown in (5) and (6) are amply attested in the OJ sources. They provide indirect evidence that Chinese § in kanbun-kundoku in fact was rendered by rna-- me 'eye', for if§ could be used to write OJ rna- -me 'eye' logographically and/ma, me/phonographically, it is because rna-- me 'eye' habitually was used to translate Chinese§ into Japanese in kanbun-kundoku. The bidirectional reading-writing relationship between kanji and Japanese words and morphemes- and what appears to be an identification of the processes of reading and writing- is evident from uses of ~}II in the earliest sources from the Nara period. In Chinese the basic meaning of ~}II (EMC *xunh) is 'instruct, teach; follow, obey', as is also reflected in many current SJ words, e.g. kunren ~}ii~Jt! 'training', kyokun ~~}II 'lesson'. It later came to be used in the sense of 'gloss, read, interpret (authoritatively)', cf. ~Jilffi!i (SJ kunko) 'exegesis, interpretation, annotation, commentary'. This is the sense and use reflected in SJ kundoku, kunten etc. In Japan ~Jil is used in this way in our earliest sources, for example in the 'reading' notes inside the main text of the Kojiki, to mean 'read (out) (a logographically written word)'. The reading notes are instructions, written in Chinese, about how to read the main text; they are not later additions, but part of the text. (7) is the first such note and

265

9.1 Kanbun-kundoku

exemplifies this usage, instructing the reader to read the kanji ;R. on this occurrence as ama, and not for example arne. (7)

r

~J'' r'i\1 ;R. :A read below say 'reading the ;R. after the (@i, say ama'

~iiJ

Ia!

•

/mal

However, a quite different use of ~}II is found in the Kojiki preface, which is generally regarded as being written in Chinese, in the paragraph outlining the writing principles employed in the main text of the Kojiki. Here ~}II is used to mean 'logographic writing', a usage not found in Chinese. The passage is generally instructive, for it explains well the tension between logographic and phonographic writing of Japanese, and we already here see the juxtaposition of lf and ~}II (SJ on, lam) which today are used about different 'readings' of individual kanji. (8) gives the text line by line together with Philippi's translation (1968: 43; emphasis added). 1 (8)

t'~, l:iliL~~, s ;i:~:tr, ~Jtt1HiJ, ~*NP;fto However, during the times of antiquity, both words and meanings were unsophisticated, and it was difficult to reduce the sentences and phrases to writing. Bjig~Jil~i!r, ~ijj/Fi~H,'o

If expressed completely in logographic writing, the words will not correspond exactly with the meaning, ~b/.lfJ!1!f' *@J!fi:o and if written entirely phonographically, the account will be much longer. ~b/.~, ~-1i];L9=J, ~Jtllf~Jik

For this reason, at times logographic and phonographic writing have been used in combination in the same phrase, ~--;Lrk.J, ~b/.~Jil~o

and at times the whole matter has been recorded logographically.

NP, 81"~[[~, b/.8:~, ;i:{1L£:(W, J!*~B:o Thus, when the purport is difficult to gather, a note has been added to make it clear; but when the meaning is easy to understand, no note is given. 1

We change Philippi's 'ideographic' to the more current 'logographic', and 'phonetic' to 'phonographic'.

266

9 The sinification of Japanese

!}f., ~tl!s

T,

~;$:/F~o

m~~~iiL ~~**'

mtHim,

:tmttz~,

Again, in the case of surnames such as Kusaka, which is written 13 T, and given names such as Tarasi, which is written 'lit, the traditional way of writing has been followed without change. In this way, alllogographic writing of Japanese derives from kanbun-kundoku reversed from reading to writing. The basic mechanism is that any character or string of characters which could be rendered into Japanese could also be used to write the Japanese rendition, as illustrated above with § writing ma- me 'eye'. Another simple example is the writing of single grammatical morphemes by single kanji, for example the use of ffiJ to write the flective -(i)te (gerund), e.g. fr{Jffi) padime-te 'beginning'. More complicated examples are the writing of morphologically complex forms, especially inflected verb forms, with logographic representation of grammatical elements, but with the order of the kanji reflecting the Chinese constituent order: (9)

/F~ ara-m 'is not' PJ~ aru besi 'should be'

erf.*P1.± sira-u-tari 'was known' Such writing is very frequent in both Norito and Senmyi5 and is also found widely in the Man 'yoshil. It is also a prominent feature of the kanji-kana majiribun way of writing which gained currency from the second half of EMJ. As mentioned above (6.1.1), Japanese writing today is a direct descendant of the kanji-kana majiribun of EMJ and LMJ, which in addition to the common principles of logographic writing of Japanese derived from kanbun-kundoku was particularly influenced by the use of kana glosses in kunten annotations. Thus, the way Japanese is written today may trace its origins directly back to kanbun-kundoku practices and kunten techniques. 9.1.2.1 Hentai kanbun~ 'kanbun' An extreme and complicated logographic way of writing Japanese has some, but not all, constituents placed in an order resembling Chinese constituent order and little specification of verbal inflection. Reading this type of text involves some of the same processes as kanbun-kundoku. (10) is a very simple example from the Kojiki which illustrates the main principles: transposing elements (here underlined) and supplying grammatical elements not represented in the text (in bold). Although it is a way of writing Japanese, this type of writing is confusingly known as hentai kanbun (~{lt>:~j: 'deviant Chinese text (or writing)')~ it is called 'deviant' because it exhibits nonChinese features (including word order and use), which is not surprising as it

267

9.1 Kanbun-kundoku

is a way of representing Japanese, not Chinese. One well-known example is the use of 1i'll to write general honorific elements in Japanese (as in (10) where it represents the honorific prefix mi-), whereas 1i'll in Chinese is used to refer to the emperor. (10)

1234567 :ONt~~.:R ~~ ®

2 so that

3 no GEN

m DAT

stma island

4 5 6 7 ama-kudari-masi-te heav en-descend-RESP-GER

1 2 3 4 56 ~:s'z:.:R;;L1i'IJ;f±

3

4

5

arne heaven

no GEN

m1-pas1ra wo HON-pillar ACC

6

2 mi-tate see-erect

1 2 3 4 5 ~:s'z:)\.~~

3 4 5 ya-piro-dono wo eight-hiro-palace ACC

2 mi-tate-tamapi-ki see-erect-RESP-SPST.CONCL

'Descending from the heavens to this island, they erected a heavenly pillar and a spacious palace' (Kojiki, Nihon koten bungaku taikei 1, pp. 52-3; translation by Philippi 1968: 50)

Hentai kanbun is subsumed as a subtype under the more general term 'kanbun', which has given rise to a great deal of confusion, because it is commonly used in a variety of meanings, to refer to quite different types of text, ranging from text written straightforwardly in Classical Chinese- this is the way we used the word above- over hentai kanbun to refer also sometimes to those portions of a Japanese text which are written in kanji. In addition to hentai kanbun, a practice arose of writing a Japanese text by reversing fully, and not just partially as with hentai kanbun, the process of kanbun-kundoku and thus so to speak translating the Japanese text into Chinese with the purpose of it being re-translated into Japanese when read. Some texts written in this way were even supplied with kunten (both kana glosses and diacritics) in order to aid the interpretation and thus have the appearance of kunten texts. It may be thought that much, if not most, 'kanbun' written in Japan since the late Heian period is not actually written in Chinese, although it looks that way, but is a cumbersome representation of Japanese. The writing

268

9 The sinification of Japanese

of Japanese 'in kanbun'- including but not limited to hentai kanbun- continued long into the modern period.

9.1. 3

Orthographic overdifferentiation

'Kun-readings', that is habitual association of individual kanji with Japanese words, have occasionally imposed orthographic distinctions on Japanese which reflect distinctions in Chinese ( cf. also 1.1.3.1 about polyvalence and equivalence in the use of kanji to write Japanese logographically). For example, the verb yom- must be written by (a) W'f or (b) :Wl< depending on whether it means (a) 'read (silently), read (out), chant (a sutra)' or (b) 'compose, write a poem', leading to the popular belief that yom- is two different words. An extreme example is the word oba (J: for 'older sister of father (or mother)' or ,f~f,J: 'younger sister of father (or mother)', imposing Chinese derived kinship differentiations on the writing of Japanese. It may well be that the split of OJ mono 'thing, being, person' into two separate words (a) 'thing' and (b) 'person' has been reinforced by the writing by two different kanji from early on: (a) ~ and (b) i!r.

9.1.4

Kokuji

A final offspring of logographic writing of Japanese is the invention in Japan of kanji for Japanese words, the so-called kokuji COO*). Well-known examples which are in use today include those in (11). Note that the final three are used with 'on-readings' to make up SJ (looking and sounding) vocabulary. 'sakaki-tree~ sacred tree'~ 3.± tsuji 'crossroad'~ m. kogarashi 'fierce wind in late autumn and early winter'~ ~ loge 'mountain pass~ peak'~ :!:Ill, ~ both hatake 'field'~ ~ hanashi 'talk'~!~ tara 'cod'~~ shitsuke 'discipline, manners'~ :it appare 'splendid, brilliant'~ fib hatarak- 'to work', do (Jj"fl rodo 'work' 2)~ ~ sen 'gland' (ff~ kansen 'sweat gland')~ ~ byo 'rivet' (il!U~ gabyo 'drawing pin').

(11)

#$ salmki

9.1.5

Kuntengo

Although kanbun-kundoku was practised in Japan long before the spread of kunten, it is not until the appearance of kunten materials that we get direct 2

j;j'IIJ was earlier written j;j'l!J, but in the Meiji period the current writing carne to be used, following a period where j;j'l!J was glossed hatarak- 'to work', which is the word Ill was made up to write. The kanji Ill itself was used already in the LMJ period.

9.1 Kanbun-kundoku

269

evidence for the language used in kanbun-kundoku, although we do have some indirect evidence from written Japanese in the form of early identifiable influence from Chinese on Japanese (through kanbun-kundoku, see 9.1.6), and in the form of early well-established associations between individual kanji and Japanese words in the writing of Japanese, as mentioned in 9.1.2. The value of such indirect evidence is difficult to judge -the risk of circularity is obvious - and should only be taken to supplement direct evidence. The language used in kunten texts is referred to as kuntengo (~JIL"#: ~). Since the beginning of the twentieth century, the study of kunten texts (see 6.2.2) and of kuntengo has become an important focus of study for Japanese linguists and philologists, more recently attracting overseas scholars as well. Kunten texts are important primary sources and it is thought that especially kunten texts from the first half of the Heian period reflect some form of vernacular language, and they thus contribute to filling the gap in our knowledge of early EMJ left by the lack of other types of sources between late OJ and c. 900. However, kunten texts are in some respects orthographically underspecified and on some points very difficult to interpret, so their evidential value is not entirely straightforward. Furthermore, strict, dogmatic norms of rendition and annotation arose, so that kunten texts from after the middle of the EMJ period generally cannot be taken to reflect contemporary language. As may be expected, kuntengo, which originates in a form of translation, is specialized and special, differing from general language in a num her of respects, in terms of style and usage. However, although kuntengo, like much translation language, most likely was stilted and in some respects formal, it first and foremost constitutes one use, or genre, of Japanese and does not differ in basic morphology or syntax from other genres of Japanese. For example, in kuntengo subject and object nouns generally have case particles to show their grammatical function, whereas subjects and objects often were left unmarked in general writing - and in spoken language. This is a difference in formality, not in grammar. Through the early Heian period, a specialization can be observed, so that some expressions were used exclusively or predominantly in kuntengo, whereas other near-synonymous forms were used in general prose writing. Apart from differences in formality and genre, this to some extent reflects that kuntengo preserves archaic features of the language. One example is the particle i (3.7.1.3) which is only attested in a small number of examples in OJ and not used at all in general writing in EMJ, but which was used extensively in some schools of kanbun-kundoku. In other cases, however, it was a matter of tendencies that some frequent features of kuntengo would not be used much in general prose writing. (12) gives some examples from the EMJ period, of which the forms in (a) reflect differences in usage or genre, whereas those in (b) reflect retention of forms which had gone out of use in the contemporary

270

9 The sinification of Japanese

language. Towards the end ofEMJ and in LMJ, kuntengo exerted a great deal of influence on fonnal writing, and some fonns were revived in writing outside of kuntengo. This coincided largely with the spread and wider usage of kanjikana majiribun (6.1.1). (12) a.

b.

9.1.6

General prose

Kuntengo

-(e)do concessive -nu, -ne, -de negative adnominal, exclamatory, gerund nar- copula

-(e)domo -zaru, -zare, -zusite

tar-

-sase- causative yafi-nar- 'be like' opase- {>owase-) 'exist.RESP'

-(a)simegotoimas-, masimas-

The influences ofkuntengo on the Japanese language

More importantly, a number of usages in kuntengo originated in the course of attempting a faithful, literal rendition from Chinese. Some such features, which arose in the translation process and which thus reflect influence from Chinese, were carried over into and gained currency in general language used outside that context, for example through dissemination of sutra commentaries or in the related didactic setsuwa literature, or through the speech and writing of scholars and clergy familiar with kanbun-kundoku. This is not unlike the kind of influence the language used in Bible translations in Europe had on European languages, where many words and idioms arose in and spread through their use in Bible translations. Examples in English from William Tyndale's translation of the New Testament (1525) include the powers that be, and eat, drink and be meny, and words such as busybody, castaway and zealous. Although it was during the EMJ period that the influence from kanbunkundoku on Japanese gained momentum and increased, it seems clear that already OJ was influenced by kanbun-kundoku. This is evident in particular in the Senmyi5 and Norito, but also in some poetry from the Man 'yoshil, but hints of this influence are found throughout the OJ text corpus. This means, conversely, that it is not possible to gauge the full extent or character of this influence, for we have no attestation of Japanese before it set in, but in the following we outline the type of influence kanbun-kundoku had on Japanese and give some examples of usage which penneated to general language. Lexical loan translations originating in kanbun-kundoku abound in Japanese through the Old and Middle Japanese periods. Early examples include ame-tuti

9.1 Kanbun-kundoku

271

'heaven and earth; the world' originating in the rendition of ~t-It!; ko no yo 'this world (opposed to before- and after-life)' < Jlt"tlt; kaku-no-goto(ku, -si) '(be(ing)) in this way'
*

272

9 The sinification of Japanese

fifty examples of kanbun-kundoku usages which are preserved in the modem language, and the great majority are such conjunctions and adverbs. A slightly different case, involving a more straightforward grammatical loan translation, is the rendition of Chinese fffi~, which is a sentence connective 'and then'. In addition to sikasite and sikaru ni just mentioned, ffiJ on its own was also rendered by the gerund formant -(i)te, and was in tum used to write -(i}te as mentioned above~ and~ 'after, afterwards' (EMC *y::1w') could be rendered by noti 'end'. The combination fffi~ was jointly rendered as (VERB)te noti which found its way into general usage in the meaning 'after VERB-ing', e.g., okosete noti 'after sending', eventually to form the model of the NJ synonymous construction VERB-te kara (tabete kara 'after eating'). A noun such as noti 'end' lends itself to grammaticalizing to acquire conjunctional uses and that may well have contributed to the acceptance into general language of -(i}te noti as a conjunctional expression, but it remains syntactically unusual in Japanese to combine a gerund and a noun in this way. A famous example of syntactic influence from Chinese, through kanbunkundoku language, on Japanese is the use of a nominalized verb form to introduce reported speech or thought, leading to a framing construction, widespread already in OJ, where reported speech is both introduced and concluded by a verb of utterance, e.g. (13). (14) is an example from the Man 'yoshil, repeated from 3.1.4.9.2 above. 'Y'

(13)

(X)

(14)

kamwiyo ywori ipi-tute-kuraku god. age ABL say-transmit-come.NMNL sworamitu yamato no soanng Yamato GEN sumyekamwi no itukusiki ruling. deity august-ACOP.ADN GEN kotodama no sakipapu bless.ADN word. spirit GEN katari-tugi ipi-tugapi-kyeri tell-continue.INF say-continue-come. STAT. CONCL

ipaku say.NMNL '(X) says "Y"'

to COMP

ipu say

kuni land kuni land kuni land

pa TOP

to COMP

'It has been recounted down through time since the age of the gods: that this land of Yamato is a land of imperial deities' stern majesty, a land blessed by the spirit of words' (Levy 1981) (MYS 5.894)

273

9.1 Kanbun-kundoku

The introduction of reported speech is atypical of Japanese, where conclusion by a complementizer and verb of utterance is usual. It is thought to have originated in kanbun-kundoku renditions of phrases like Chinese ~ 8 'the master says: ... ', introducing sayings by Confucius in the Analects, which in kanbun-kundoku typically is glossed as si ipaku (>mid EMJ iwaku), using the nominal form of ip- 'say' (> iw-). The rendition of~ 8 'X' as si ipaku X' (to ipu), rather than for example si X' to ipu, maintains the original word order, and again information structure, and keeps the rendition of~ and 8 together, at the expense of creating an unusual sentence construction. From EMJ, the nominal verb form disappeared from general use, but this construction continued to be used through the MJ period with other nominalizers, e.g. koto or SJ yafi 'way, manner', as in (15). It is no longer used productively in the modem language.

(15)

kaditori oarsman

no GEN

ipu say

kurotori (black)scoter to COMP

zo FOC

yafi manner no moto ni siroki nami wo yosu GEN base DAT white wave ACC break

tpu say

'The oarsman said: "below the black birds, the white waves are breaking" ' (Tosa)

Although the nominal form ceased to be productive and dropped from general use, a number of such forms continued to be used in kanbun-kundoku in these constructions, found their way into general language as sentence initial adverbs or nouns, and are retained into the modem language: iwaku mentioned above, which is now used as a noun 'reason, pretext; past' and also to introduce quotes or proverbs ('as the saying goes'); negawaku wa 'I pray, hope'(< OJ negapaku pa 'pray.NMNL TOP; what I pray'); omoeraku 'methinks' (< omopyeraku 'think-STAT.NMNL; what I am thinking'); omowaku 'thought, opinion' (today written }i!t~) ( < omopaku 'think.NMNL; what I think'); osimuraku wa 'regrettably, unfortunately' (< wosimuraku pa 'regret. NMNL TOP; what I regret'); osoraku 'likely, probably' ( < oso(m)raku 'what I fear'). Other examples of usage retained in kanbun-kundoku language and subsequently carried over into general language include the OJ passive -(a)ye-, which dropped out of the language in the transition from OJ to EMJ, but is reflected in lexicalized modifiers such as iwayum 'so-called'(< OJ ipa-yum 'say-PASS.ADN'), arayum 'all, every' (< ara-yum 'exist-PASS.ADN'), both of

274

9 The sinification of Japanese

which are in use today. In kanbun-kundoku, ipayuro and arayuro were used in rendition of phrases such as errffi!i NOUN and err1:E NOUN, respectively. In Chinese err (EMC *~i:~') is both a noun 'place' and also, as in these examples, a subordinator used in some relative clauses, and in these two cases err was rendered by adnominal verb forms, which subsequently passed into general language as lexicalized forms. Interestingly, a more literal way of rendering Chinese errvERBZNOUN ('NOUN which VERBs'), instead of simply forming a normal Japanese relative clause, arose in kanbun-kundoku and was carried over into written language in the form VERB tokoro-no NOUN, where tokoro-no functions as a complementizer between the relative clause and the head noun, for example (16); see also 12.6.1.1.2. This usage is clearly motivated by the kanbun-kundoku rendition of err as tokoro 'place, and as no, combined with the use of Chinese err in relative constructions. As with the quotative framing construction mentioned above, this construction is not ungrammatical in Japanese, but nor is it motivated internally within Japanese. In the Meiji period this usage was revived in kanbun-kundoku-like translations of relative pronouns in Dutch.

z

(16)

tatekome-taru tokoro-no to close-STAT.ADN door 'The door which had been closed' (Take tori)

9.2

Ondoku

Ondoku (l!fW'E 'pronunciation reading') is the reading and vocalization of Chinese text in Chinese, learned as a foreign language, without rendition or translation into Japanese. Over time a great many loanwords have entered Japanese based on this way of reading Chinese texts. Today somewhere between thirty-five and sixty per cent of words in running text, depending on genre, are SJ loanwords, and it is customary to speak of a distinct SJ vocabulary layer in the Japanese lexicon. The term 'Sino-Japanese' is ambiguous and that has given rise to several misunderstandings. There are three distinct, but interrelated issues, which are not usually distinguished explicitly: (a) JapanoChinese: Chinese as a foreign reading language in Japan (9.2.1); (b) SinoJapanese: nativized norms for pronouncing kanji (9.2.2); (c) Sino-Japanese loanwords: loanwords in Japanese deriving from J-Ch or SJ (9.2.3).

9.2.1

Chinese as a foreign (reading) language; Japano-Chinese

In the initial period of contact with Chinese text and language, the fifth and sixth centuries, this contact is thought to have been indirect and primarily to

9.2 Ondoku

275

have been mediated by immigrant or visiting scholars and later monks and nuns from the Korean peninsula, probably especially from the kingdom of Paekche, who taught and expounded Chinese texts: first Chinese classics, and later, after the introduction sometime in the sixth century of Buddhism into Japan, also Buddhist sutras and commentaries. This was done in part through kanbun-kundoku, but also through and in Chinese. An important element of studying texts in Chinese was the enunciation and vocalization of text, particularly in the recitation of Buddhist sutras. For that reason attention is often focused on the pronunciation of Chinese, hence the term ondoku. For lack of a more elegant term, we will refer to Chinese used in the study and reading of Chinese texts in Japan, that is to say, Chinese employed in ondoku, as 'Japano-Chinese', emphasizing that it was a variety of Chinese language. It is often thought that the Chinese taught and learned in Japan in this early period was based on southern Chinese varieties, but this is not really known. It seems clear (a) that it does not reflect any single variety of Chinese, but is a cumulative and multi-layered conglomerate of varieties of Chinese, most likely transmitted by scholars and monks from the Korean peninsula in the main in the fifth through early seventh centuries; and (b) that it is not a single uniform norm, but exhibited variation between different schools and sects within which conventions for reading and reciting texts, especially sutras, became fixed, with orally transmitted pronunciation norms, which gradually became increasingly removed from Chinese spoken in China. However, through the seventh and eighth centuries there was extensive direct contact with Tang China, with envoys and students dispatched to visit and study in China. In the course of their studies they acquired contemporary Chinese as spoken in the Tang capital Chang'an and brought this back with them to Japan, introducing new, competing pronunciation norms and readings of texts. This led to tension between the old and the new ways of pronouncing Chinese. Between 792 and 806 (during the reign of emperor Kanmu (737-806, r. 781-806)), several imperial decrees were issued that the proper Chinese pronunciation (~i!f Japanese kan-on 'Han(= Chinese) pronunciation' or lE i!f sei-on 'correct pronunciation') be used, both in the study and reading of Chinese classics, which by then had become all important for civil service exams, and in official and public recitations of Buddhist sutras. The decrees used phrases such as ~-%nf~i!f; o/JJt.l ~i!f 'make everybody read in Chinese pronunciation; do not use Wu pronunciation' (cited from Yuzawa 1996: 47). The older established norms of J-Ch have variously been referred to as fiJi!f (wa-on 'Japanese pronunciation'), ~t~i!f (tsushima-on 'Tsushima pronunciation' reflecting an early putative route of transmission of Chinese), or ~i!f (go-on 'Wu-pronunciation'), which is often interpreted as showing that early J-Ch was based on south-eastern Chinese from the region around present-day

276

9 The sinification of Japanese

Shanghai which was the seat of the ancient, barbarian Kingdom of Wu. We here adopt the most widely used terms, go-on for the older (composite) norm(s) and kan-on for the new norm. The necessity of issuing decrees shows that the replacement of the older well-established pronunciation norms did not proceed without resistance, which was particularly strong from Buddhist sects to whom the familiar, wellestablished vocalization of sacred texts was understandably important. This may be compared to the widespread resistance against reformations of Bible translations (or even earlier against translating the Bible into vernacular languages) in Europe. Consequently, the older pronunciation norms were retained in many Buddhist sects and traditions and also used for some sutras introduced after the issue of the decrees. It is worth emphasizing that these decrees concerned J-Ch, that is to say, the norms for reading Chinese text(s)~ they did not concern the shape and pronunciation of SJ loanwords which at the time had already been taken into the language. Kan-on was at first used both for study and reading of text and for practical communication. It was maintained both by instruction by native speakers, through regular intercourse with China (also after the official envoys to China ceased towards the end of the ninth century), and by study of Chinese rhyme tables and pronunciation guides. Efforts were made to maintain correct standards of pronunciation, and, as in China, the correct, normative 'reading' of individual characters became an independent object of study. As a result, kan-on in the course of the ninth and tenth centuries became a fossilized norm and over time as removed from Chinese spoken in China as the older go-on had been. Later, competing ways of pronouncing Chinese, representing different varieties of Chinese with different phonetics and phonological categories, came to Japan, especially in the context of Japanese Zen Buddhism from the late twelfth century onwards, said to be based on southern Chinese varieties, leading to the establishment of J-Ch pronunciation norms within some Zen sects which were different from the kan-on, referred to as so-on ('Song pronunciation', *if, after the Song dynasty, 960-1279) or more commonly as to-on ('Tang pronunciation', f,!f{f), which we use here. Note that to f,!f was used in the sense of '(real) Chinese', like kan ~had been centuries earlier when the kan-on pronunciation was being promulgated as the real Chinese pronunciation, and that it does not refer to the Tang dynasty (618-907). Thus, three different main varieties of J-Ch can be distinguished, defined in relation to kan-on: (a) pre-kan-on comprises the varieties collectively referred to as go-on and sometimes wa-on~ (b) kan-on is the mainstream norm which arose out of J-Ch based on the Chinese of Chang'an in the late Tang period~ finally, (c) post-kan-on are the subsequent to-on pronunciation norms. Although we speak of three main varieties, there must have been considerable variation within them, depending on the traditions of individual schools and Buddhist sects.

9.2 Ondoku

277

9.2.1.1 The phonetics and phonology ofJapano-Chinese The several varieties of J-Ch have played important parts in the history of Japanese: each has contributed loanwords to Japanese, and go-on and kan-on formed part of the basis for determining the choice and sound values of ongana (1.1.2.5). In essence, however, J-Ch varieties are varieties of Chinese -even if they were functionally restricted- and the study of phonetics and phonology of J-Ch is best thought of as a branch of Chinese historical phonology, of less concrete importance for the study of Japanese. Apart from reflexes in Japanese loanwords or in the sound values of man 'yogana, we know little about the phonetics of J-Ch. It is inevitable that some amount of phonological approximation to Japanese took place in the course of acquisition and over time as contact with the source language(s) was lost, and perhaps also as a result of wider popular participation in sutra recitation. It is, however, also clear that J-Ch sound systems were foreign, deriving and maintaining their prestige in part from being identifiably foreign in sound texture. An illustrative analogy might be the pronunciation of French taught in English schools (until recently, at least) which is elaborate, incorporates stereotypical and often exaggerated features of French phonetics, but also is characterized by a great amount of phonetic approximation to English~ which in short sounds neither English nor French. The most important material we have to study the phonetics and phonology of J-Ch are pronunciation glosses and glossaries for individual texts and dictionaries (see 6.2. 3), which give instructions for the pronunciation of characters or words. However, much of this material is unpublished or not easily accessible, and in any case it is difficult to interpret, for (in addition toflinqie glosses, see 6.1.4) it employs man 'yogana or kana. Generally the material is interpreted and presented in terms of the orthographic categories of the kana syllabary, which was used to record Japanese and whose orthographic categories reflect Japanese phonology. This makes it very difficult to acquire knowledge about differences between Japanese and J-Ch. A basic question is: Did the phonograms used in pronunciation glosses for J-Ch have the same reference (sound values) as when used in writing Japanese? This issue is further confounded by the fact that glosses in texts or dictionaries are often cited today as if they give information about the pronunciation of SJ loanwords or of individual characters used in SJ loanwords, rather than as pronunciation guides to J-Ch. In addition to glosses, living traditions of sutra recitation potentially hold important clues to the phonetics and phonology of earlier J-Ch, although they are said not to constitute unbroken traditions, but over time to have been subject to normative correction in terms of SJ kanji readings. 9.2.1.2 Early Middle Chinese, Japano-Chinese and ongana By 'Early Middle Chinese' (EMC) we refer to the Chinese language reflected in the Qieyitn (-glJf\i), a dictionary which by means offiinqie glosses (cf. 6.1.4)

278

9 The sinification of Japanese

records the correct pronunciation of Chinese characters. It was compiled in the late sixth century by a group of scholars from different parts of China and published in 601 by Lu Fayan; although there are no extant copies, it is thought to be faithfully reflected in the eleventh-century dictionary Guiingyim (lt:~). The pronunciations recorded in the Qieyiln seem to be an overdifferentiating compromise between the scholars involved, who were also not quite sober when they tried to agree on the pronunciation of individual characters. The phonology of the Qieyun thus does not represent any one variety of early Chinese, although it is often taken to reflect the language of Chang'an. However, EMC is well reconstructed, and it therefore constitutes a convenient reference point form ost of Chinese historical phonology, and also for studies of the sound values of man 'yogana. Even so, it is important to keep in mind that none of the varieties of J-Ch directly reflects EMC, as kan-on postdates EMC and go-on both predates EMC and may be based on a somewhat different variety of Chinese. Ongana were written representations of OJ sounds. The ongana were to a large extent chosen to represent OJ sounds on the basis of a perceived similarity between the OJ sound and the J-Ch pronunciation associated with a kanji, although we saw above that use of some ongana must have been based on other criteria than their sound values in J-Ch, for example the use of 3'E to represent OJ !myel (cf. 1.1.2.5). All of this shows that the relation between the reconstructed EMC sound values and and the sound values of the ongana was very indirect.

9.2.2

Sino.Japanese

As mentioned, the distinction between J-Ch and SJ is not a common one and SJ is usually equated with J-Ch, as if they were one and the same thing. However, they are fundamentally different: SJ is a nativization of J-Ch, removing it from the realm of a foreign language and providing a nativized pronunciation norm of kanji, which derives from J-Ch, but which in contrast to J-Ch is in full conformity with Japanese phonology and can be used within Japanese. SJ is not a language, less so than J-Ch, but essentially a norm for pronouncing kanji. It may in many respects be likened to latinate English. SJ made available for easy vocalization in Japanese the full range of words from Chinese. SJ was used primarily in ondoku, that is reading out Chinese text without rendering it into Japanese, but it is not the case that SJ has replaced J-Ch. For example, whereas ondoku of most Chinese texts today is no longer done in J-Ch but in SJ, some schools of sutra recitation even today maintain the use of J-Ch, and that was much more so through the pre-modem period. As SJ derives from J-Ch, different layers of SJ resulted, reflecting the three main varieties of J-Ch, and we here use the same names for them, speaking of SJ go-on, SJ kan-on and SJ to-on.

9.2 Ondoku

279

As the distinction between J-Ch and SJ is not usually made explicitly, it is not possible to say much in detail about the history and development of SJ. The common term for what we today call ondoku seems earlier to have been kowe (>late EMJ koe) 'voice', whereas kundoku was mostly referred to by yomi 'reading', although kuni or kun (~}II) were also used. There are several references in both EMJ and LMJ texts to reading out of Chinese text or words, using the term kowe 'voice'> koe, but it is not clear whether they refer to J-Ch or to SJ. From early LMJ we find references to different types of on (if), variously named but reflecting the three layers of J-Ch and SJ. It seems likely that nativizations of J-Ch to give SJ became established towards the very end of the EMJ period when we see an increase in the use of SJ vocabulary written in kanji, in the kanji-kana majiribun texts. The greater variability in sound shapes and even morphology of SJ loanwords until the second half of EMJ (see 9.2.3.2) indicates that loans from the first half of EMJ were taken in from J-Ch, but mostly from SJ thereafter, and this lends further support to this dating of the establishment of SJ. The existence of SJ was also a prerequisite for the occasional emergence of SJ words by using the SJ readings of kanji which were used logographically to write native words (see 9.2.4) and this is something we find from LMJ. For example, in early LMJ the written form 1ffi!IDJ, which was used to write EMJ omae (< OJ opomapye), gave rise to gozen by using the SJ reading; and that would not have been possible without a SJ reading norm. At the end of LMJ, Rodrigues in Arte da lingoa de Iapam (see 10.2.2.2) makes reference to what are clearly three layers of SJ, 'coye', i.e. koe. He exemplifies different readings of kanji, citing 'govon' (go-on), 'canvon' (kan-on), and 'toin' (to-on), of for example 1f (guio !gy:J:JI, co lb:JI, an) which since then have been used as the stock example in most exemplifications of the three different types of readings. His examples and explanations make clear that the notion of 'coye' first of all pertains to ondoku. He thus gives by way of illustration five different versions of the name of the Lotus Sutra (SanskritSad-dharma-pUfl~arfka-sutram 'truelaw-lotus-sutra, i.e. sutra of the lotus of the true law/teaching', which in Chinese is rendered WJ;¥~1-1~*£) (Arte, p. 666, following Doi's emendations): in (17a) 'bongo' (Sanskrit; an adapted pronunciation which is not very far from the original Sanskrit)), (b) 'govon', (c) 'canvon', (d) 'toin', and (e) 'vago' (Japanese, i.e. a kanbun-kundoku rendition of the Chinese, here glossed in (18)). Thus, at the very end of LMJ there were three well-established layers of SJ, known by the same names we use today and used in different traditions of ondoku. Note, however, also that they were thought of as different languages, on a par with Sanskrit and Japanese; and finally that the SJ version of the name of the Lotus Sutra in use today (Myi5hi5renge-kyi5) is the one identified as go-on. It is interesting that ikada 'raft (/small boat)' is used to render Chinese*!, which translates Sanskrit sutra 'thread, string, rule'.

280

9 The sinification of Japanese

(17) a. Satarama fundariquia sotaran. b. Me8 f8 rengue qiuo (/myoo foo renge kyoo/) c. Be8 fo renga quei. (/byoo foo renga kei/) d. Be8 fa renga quin. (/byoo fa renga kin/) e. Taye naruya norino fachisuno fana icada. (18)

tae naru ya wonderful is EMPH

non no law GEN

fatisu no lotus GEN

fana ikada flower raft

9.2.2.1 Present-day Sino-Japanese Present-day dictionaries of kanji as used in Japan and in Japanese, so-called kan-wa (~fiJ) dictionaries, give for almost all characters listed in them one or inmost cases several SJ readings, 'on-readings', often described as 'Chinese readings', or readings which derive from Chinese. SJ readings are usually classified into four overall classes, which are defined in reference to the three main varieties of J-Ch: go-on, kan-on, to-on, and kan yo-on ('idiomatic readings' 'tltJt.llf, readings in common use, which do not conform to regular assignment to the other classes). The SJ readings listed in present-day dictionaries are an intriguing conglomerate of prescriptive readings and of actual recorded usage, observed in SJ loanwords and in SJ reading traditions. The prescriptive SJ readings in presentday dictionaries were to a large extent established deductively by philologists (such as Moto' ori Norinaga) in the Edo period through study of the Chinese rhyme tables, in particular by application of the phonological categories of the Yimjz'ng (rul~, Japanese inkyo). The Yimjz'ng is thought to reflect the pronunciation of the second half of the Tang period ( 618-907), i.e. largely the Chinese on which kan-on J-Ch was based, and itself to date from the late ninth century or the first half of the tenth century. The Yunjz'ng came to Japan in the thirteenth century and since then came to form an important basis for the study of kanji readings, first in J-Ch, but later also for the study and codification of SJ kanji readings. The effort in the Edo period aimed at establishing correct SJ reading and pronunciation norms, as well as norms for correct etymological spellings of the readings. Almost all characters are assigned a kan-on and most also a go-on reading which is constructed as a 'correct' pronunciation in correspondence with the categories in the Yunjing. These readings thus need never have been used in SJ or in a SJ loanword, but nor were they ever intended to provide a record of current readings. On the other hand, the SJ readings listed in dictionaries also do include observed usage. While it would be easy enough to construct 'correct' to-on readings for all kanji, to-on readings included in dictionaries are generally those attested in SJ loanwords. And the kan yo-on listed in dictionaries are by

9.2 Ondoku

281

Table 9.1. Kanji readings EMC

Go-on

Kan-on

11' i!l'

to;;"iajk *yaijlJ(h) /y e:jl)<~ *y:njlJ' /ye:jl)'

:IT

*t~

"*

*mit

syaku (shaku) gyaii(gy6) gyail (gy6) toil (to) tyail (cho) miti (michi) mitu (mit su) si (shi) soil (so) .syail (sho) zyail (jo) netu (netsu) neti (nechi) su zyaku (jaku)

seki kaii (ko) kail (ko) toil (to) tel (lei) bitu (bitsu)

.R

:::;.. ~

*tsi' /tsi' * ~aiW!JI ~reiWI) *dziajlJ

W,\

*Jliat

~

•

*fj:U~

:£"

*d~ajk

~

*JlUWk *Jluawk

11il IJil

*miaw *miajlJ *lip *pEj

g

ft ~

"'lt t.l :flll

*t~:njtj t~ :jl)

~

*kEjlJ *kiajlJ

*4'J12!! ~ ~ ~ ~

1"1

*~w *tt~ip/ t~hE:p

*p~it

*tajlJ' *t~k

*d~alJ

*•it *?iajlJ *truwk

niku niku noku meu (myo) myail (my6) ripu (ryii) pai (hai) .syail (sho) syu (shu) sepu (sho) kyail (kyo) kyail (kyo) pili (hi chi) tyail (cho) toku zyail (jo) ziti (jichi) yail (yo) tiku (chiku)

si (shi) sail (so) sel (sei)

To-on

Kanyo-on

an an

kyail (kyo)

ton tin (chin)

su sin (shin)

zetu (zetsu) .!Jill (shu) seki

yu koku syaku (shaku)

ziku (jiku) zyoku (joku) beu (by6) me"i (mei) ripu (ryii) pei (hei) sail (so) sou (so) sapu (so) ke"i (kei) ke"i (kei) pitu (hitsu) tel (lei) toku syail (sho) situ (shitsu) e"i (ei) tiku (chiku)

min ritu (ritsu) pi (hi)

kin kin pilei (hiki) da

zitu (jitsu) situ= /siQ-/ (shitsu)

definition (part of) the sound shapes of SJ words which do not confonn to the nonnative go-on or kan-on, and are not to-on. It should finally be noted that not all SJ loanword shapes are captured by the SJ readings in dictionaries. That applies especially to SJ loanwords from the Heian period (see 9.2.3.2). Table 9.1 gives some examples of SJ kanji readings classified in this way, giving for reference also the reconstructed EMC readings, and in 9.2.2.2-3 we

282

9 The sinification of Japanese Table 9.2 EMC syllable initial consonants

EMC

Go-on

Kon-on

*m

m

*m(V~)

m

b m

*n

n

*n(V~)

n

d n

*b *p

b p

p p

*d *t

d

*g *k

g k

*z *s

z

k k

give a brief overview of some of the commonly observed correspondences between go-on, kan-on and EMC. As mentioned above, it must be emphasized that none of the layers of J-Ch or of the SJ character readings directly derives from EMC, but with this caveat in mind, we will say that SJ readings 'reflect' EMC. Although the phonological adaptation of SJ loanwords in the EMJ and LMJ periods was not as regular as these correspondences suggest, they do to a large extent also hold for SJ loanwords, and for that reason we give the SJ readings in the sound shape we reconstruct for EMJ on the basis of their historical spelling and the EMC sound value. This is strictly speaking anachronistic, as many readings date from NJ, but this will give an impression of the sound changes undergone by SJ loanwords.

9.2.2.2 Syllable initial consonants EMC seems to have had thirty-eight different syllable initials (reconstructed by Pulleyblank as /p, ph, b, m, w; ~, t~h, d~ ~, ~ ts, tsh, dz, s, z; tr, trh, dr, nr; t, th, d, n, 1; tc;;, tc;;h, ~, J1, c;;, ~, j; k, kh, g, IJ, x, y; ?/) which is a somewhat more complex system than Japanese. Only some are given here as representative. Go-on nasals correspond to kan-on mediae, except usually when followed by -Vi1 or -Viwhich are the reflexes ofEMC *-VIJ, and go-on mediae correspond to kan-on tenues. Regular sound changes have changed /p-/ > NJ !h-1 (7.3.1, 11.3, 14.3) and /d-/ to lz-1 before li, u! (14.1). 9. 2. 2. 3 Syllable final consonants EMC had six main syllable final consonants (/p, t, k, m, n, IJ/) with fairly straightforward correspondences. Go-on and kan-on largely agree in their

283

9.2 Ondoku Table 9.3 EMC syllable final consonants EMC

Go-on

Kon-on

*-p

pu

pu

*-t

ti

tu (t)

*-k

ku I {e, a, o, u} _ ki I {i} _

ku I {i, a, o, u} _ ki I {e} _

*-m

N

N

*-n

N

N

*-1)

fi I {i, a, o, u} _

fi I {i, a, o, u} _ I/e

II e

reflexes of syllable final consonants, except with EMC *-t, *-k. Regular sound change has changed /-pu/ > 1-ul (7.3.1.1) and 10, I/> /u, if (11.1.2). As shown in (Table 9.3), EMC *-tis usually said to be reflected in go-on -ti and kan-on -tu, but in fact, at the end of the LMJ period most kan-on and some go-on reflexes of EMC *-t were simply 1-t/, which mostly became /tu/ (sometimes /til) in the NJ period. Syllable final /-t/ and the reflexes in SJ vocabulary of EMC *-twill be discussed further in 11.4; see also 7.2. EMC *-IJ was reflected differently, as /U/ or li/ depending on the preceding vowel. Combined with a difference between go-on and kan-on in the reflex of the vowels and glides in EMC *(C)iaj(IJ) as go-on /yal and kan-on lei (realized with a phonetic onglide or strong palatalization of the preceding consonant: [ie] or [qe], cf. 2.3, 7.3.2.4), this has resulted in a frequent correspondence between go-on and kan-on: go-on EMJ (C}yai1 > LMJ (C)y:x1 > NJ (C)yoo > cNJ (C)yo :: kan-on EMJ (C)e'f> NJ (C)ei. For example:~ yo:: ei, 1:. sho :: sei, Jjt kyo :: kei. Many readings reflecting EMC *-p, *-t, *-k have variants in final /Q/, which are not usually listed as separate readings. They result from assimilatory processes in combinations with a following consonant: -pu(-C-) > -Q(-C-); -ti(-C-)1-t(-C-) > -Q(-C-); -ki(-k-)1-ku(-k-) > -Q(-k-). Finally, there is a small group of kanji with readings where EMC *-p is reflected as -tu > cNJ -tsu, e.g. ""fl. ritsu (as in cNJ ~!R""fl. dokuritsu 'independence', cf. EMC *lip) or* zatsu (1i*fokuzatsu 'complicated', cf. EMC *dZlp/dzap). 3 These readings are thought to have arisen as back formations from assimilated forms such as those above, with the final -Q analogically interpreted as deriving from -tu. Like the kan-on reflexes of EMC *-t, these forms had final -tat the end of 3

The relevant kanji are (giving in brackets the kan'yo-on and the EMC sound values): ft (Jean yo-on: ritsu; EMC *lip), ~ (zatsu; *dz:~pldzap), ~ (.retsu; *tsiap), ~ (setsu; *9iap), ~ (shitsu; *t~;ip), ~ (shitsu; *~;ip), I} (chitsu; *drip), ;m (satu; *s:~plsap); see Komatsu (1956).

284

9 The sinification of Japanese

LMJ (11.4), suggesting that the back fonnation was from -Q to -t which later became -tu, that is: -pu-(C-} > -Q-(C-} > -t > -tu.

9.2.3

Sino-Japanese loanwords

A number of identifiable Chinese loanwords were used already in OJ (4.2.2). Some of them are undoubtedly very old, predating the study of Chinese text, and it is likely that some entered Japanese through the Korean peninsula. However, that early trickling of loanwords from Chinese is insignificant in comparison with the great number of loanwords which entered the language through the medium of J-Ch or SJ, by being used outside the context of reading Chinese text and eventually carried over into everyday language. It is such words we refer to as 'SJ loanwords'. As mentioned above, the distinction between J-Ch and SJ is not usually made explicitly and it is therefore difficult to detennine for many loanwords whether their proximate source is J-Ch or SJ. However, it is noteworthy that among SJ loanwords until the second half of EMJ, we see greater variability in and adaptation of sound shapes than later, and we also have small but noticeable numbers of examples of morphological adaptation where SJ loanwords become verbs or adjectives (9.2.3.2), whereas that is much rarer later. All of this indicates that loans from the first half of EMJ were taken in from J-Ch, whereas most loans from the end of EMJ onwards were taken in from the nativized SJ. Through time many SJ loanwords must have entered the language through writing, but it is clear that the majority of SJ loanwords, especially in the first half of the EMJ period, had gained currency in everyday spoken language. However, it is equally clear that there was some consciousness about their provenance, for example because they were successfully avoided in poetry; that would not have been possible if they had been fully assimilated. As with loanwords from European languages centuries later, SJ loanwords had, and today to a large extent still have, a socio-linguistic status well captured by Gennanfremdwort ('alien word'; cf. also 17.1.1). The intake of SJ loanwords was significant enough to affect phonology, especially phonotactics (7.2), and to some extent grammar, especially word classes (8.1.3).

9.2.3.1 Sino-Japanese loanwords in Old Japanese OJ is thought not to have had many SJ loanwords. Very few are found in the OJ poetic texts: in the Man 'yoshil a small number of SJ loanwords is thought to be found in a series of poems in volume 16 (3827-58), see (19). Many of them are logographically written, so it is not always clear that they in fact are SJ loanwords and not rather logographically written native words. It is on the other hand also possible that some of them are not SJ loanwords, but (Japano-)

9.2 Ondoku

285

Chinese words used for stylistic effect. In any case, the phonological shape of the logographically written words is guesswork. In the first poem in that series (MYS 16.3827) the kanji - =~IZ!lli:A appear, but there is no agreement about whether they were intended to represent SJ or native Japanese numerals; Ji and :A appear elsewhere in these poems, clearly representing SJ numerals and traditionally read as go and roku.

(19)

~~~~ baramoni 'Brahman' (EMC *pa la m~n, Skt. Briihmarza).

This is the traditional reading of the ~~ in the single occurrence of this word in the Man 'yoshii; in EMJ a more regular SJ version came to be used: baramon, which is the form the word has today. ;fl~ daniwotildanawoti 'benefactor' (EMC *dan wuat, Skt. diinapati 'benefactor; giver; giving-lord'). The two versions are the traditional readings of the single occurrence in the Man 'yoshii; in EMJ a more regular SJ, dan. wotu, came to be used, but the nativized version survived in a reduced form, danna 'benefactor', eventually giving NJ danna 'husband, master'. Ji{lz: gowi 'fifth rank'. ~ kaii 'fragrance' (in some reading traditions read kori 'fragrance' which is a native word or a naturalized loan). :;IJ kuii 'accomplishment, merit'. ~{PJ~(-JJ-~) mugau(-no-satwo) 'Mugo village; not-evenanything village, village of nothingness' (legendary place of natural emptiness, void of human artifacts; (both this and the following are from Chuang-Tzu, a Chinese Taoist classic (fourth or third century BC) which exerted great influence on later Chinese Buddhism); ~~iSM(-t~- Jl.J) pakwoya(-no-yama) 'Mt. Hakoya' (legendary dwelling of sage hermits). ¥~~ffi and{~ popusi '(Buddhist) monk, priest' (in the Nihon shoki r~~ffi is glossed as popusi). .1J±(.) rikizi(-mapi) 'strong-man(-dance)' Buddhist ritual dance. ~~ saukepu 'Saikachi' (name of a tree; some reading traditions have pudi-no-kwi, kapara-pudi). 1:.:71: syaiizi 'living and dying' (in some traditons read ikizini/ikisini which is a loan translation of 1:.:71:, cf. 9.1.6). ~ tapu 'stupa' (the Chinese is a loan from Skt. stiipa).

286

9 The sinification of Japanese

Outside that series of poems, only ~ffl kwaso (15.3754 'travel pass') and ljiJJ ~ teusan (18 .4121 'coming to court' (other readings include mawiri, mikadwomawiri, miyade)) are found in the Man 'yoshii. By contrast, the Senmyo had a significant num her of loanwords from Chinese: Vovin 2005: 60-2 lists sixty-six SJ loanwords found in the Senmyo; they are all written logographically and we do not know their sound shape. Many of these words are fairly specialized terms (~rt 'protect the Buddhist law', ~!ri 'bodhisattva', 'office manager', )l~:fff 'obedient grandchild') and it is quite possible that some of these were J-Ch words used within Japanese, but others are more common words which probably were established loanwords (~ 'music', ~R 'rebellion, treason', ~ 'sutra', tFf 'beheading'). Taken together with the fact that SJ loanwords were normatively excluded from poetry in EMJ, this leads to a questioning of the usual assumption of absence to any significant extent of SJ loanwords in OJ: It is quite likely that their scarcity in the written sources from OJ reflects that this norm for poetry was already active in OJ, but that at least learned or official language had far more SJ loanwords than is usually thought.

:m*

9.2.3.2 Sino.Japanese loanwords in Early Middle Japanese Also in the Japanese poetry from the EMJ (and LMJ) periods we find almost no SJ loanwords. As mentioned earlier, this absence from poetry confirms their sociolinguistic status as identifiable fremdworter. In prose writing, however, the use of SJ loanwords can be seen to increase steadily through the period. The following word counts show the proportion of SJ loanwords in some well known sources (extracted from Tsukishima 1969: 588-9, 1987: 277-8).

(20)

Text frequency of SJ loanwords(% of words in running text) Ise monogatari Tosa nikki Kokin wakashii (kana preface) Genji monogatari Hamamatsu chiinagon monogatari Daijionji sanzohoshi-den

(21)

6.2% 4.0% 11.3% 12.6% 14.3% 43.0%

Lexical frequency of SJ loanwords(% of different words) Genji monogatari Konjaku monogatari Sangoshiiki-chii Daijionji sanzohOshi-den

4.8% 14.0% >50.0% 85.8%

9.2 Ondoku

287

In the prose written mainly in hiragana, the proportion of SJ loans out of distinct words remains well under ten per cent through the period, but it is somewhat higher in kanji-kana majiribun, as for example in the Konjaku monogatari. This is related both to orthography and subject matter: First, as such texts include a high proportion of logographic writing it was easy to use SJ words or phrases written in kanji. Second, kanji-kana majiribun was often setsuwa literature about or inspired by Buddhism, and it included much Buddhist terminology and many Indian or Chinese proper names. Thus in the Konjaku monogatari the stories taking place in India and China which have a strong Buddhist element have a much higher proportion of SJ loanwords, whereas in the stories which take place in Japan, the proportion of SJ loanwords is more or less as in the Genji monogatari. It is not surprising that some kanbun-kundoku texts, such as the Daijionji sanzohOshi-den annotations, contain a large proportion of words which were left untranslated as either SJ loanwords or J-Ch words. There is, to be sure, much vocabulary relating to Buddhism and politics and philosophy among the SJ loanwords, in addition to words relating to the life, positions and ranks at the sinified court, conforming to the picture of the use of SJ or Chinese words in OJ. However, the variety and number of everyday vocabulary, including emotional and expressive vocabulary, is striking, showing that a significant num her of SJ loanwords had become well integrated into everyday language, at least of the court nobility and educated classes. Another remarkable feature is that in terms of SJ loanwords the vocabulary composition ofEMJ does not differ much from that ofNJ, apart from frequencies of usage. Morphologically, SJ loanwords were invariably taken in as nouns, but in addition to general nouns (22), we already at the beginning of EMJ have a well populated class of verbal nouns used with the light verb se'do' (23), and a sizeable class of adjectival nouns/adverbs used with forms of the copula (ni infinitive, no adnominal, to infinitive, or extended ni ar- or, more rarely, to ar-), see (24). (22)

Nouns sauyaku (lit~) 'a herbal medicine'; daitoko (:*1~) 'monk of great virtue'; inmyafrzi (~~Gfli) 'fortuneteller', nikki (El~c) 'diary', sipu (#:)'anthology', neti (~)'fever'

(23)

Verbal nouns gu-se- (A.) 'furnish, be furnished (with)'; si-se- (:7E) 'die'; wen-ze- (!&) 'resent'; ron-ze- (~) 'discuss'; rongi-se- (~~) 'debate'; penge-se- (~1~) 'transform'; kuyail-ze- (~:i€) 'make offerings and prayers'; buku-se- (H~) 'drink'; kei-se- (~ 'speak HUM'

288

9 The sinification of Japanese

(24)

Adjectival nouns/adverbs gokuneti-no 'very hot' (fi~)~ pizaii-nari 'extreme'(~~~) mail-nilno 'strong, fierce' (~) toni-ni 'swiftly, suddenly'(~)~ ziti-ni 'really' (~)~ beti-nilno 'particular(ly)' (JJIJ)~ gen-ni 'really' (ffl.)~ seti-ni 'definitely' (-glJ)~ zinen-ni (§t'~) 'spontaneously'~ sizen-ni (§t'~) 'of, by itself' rinrin-to 'very cold' (~i~)

We also find compounds of SJ and native words, see (25), in some cases with some phonological integration, as well as native grammatical material used with SJ words, see (26). Borrowed SJ prefixes, such as ko- 'late, deceased' (~), sai- 'most' (Ji), combined with native lexical material. Also, derivational morphology was borrowed from SJ, e.g. -yail '-like' (~) which derives an adjectival noun from a noun (cf. 8.6). Finally, we see a few examples of SJ loanwords used as inflected adjectives, (27), or even verbs, (28)~ these are few in number and may to some extent be the result of literary playfulness and inventiveness, but they do conform to well-established native morphological patterns, such as reduplication to form shiku adjectives. All of this shows that SJ loanwords formed a well established and integrated part of the EMJ lexicon, even if still recognizable as fremdworter. The examples of SJ loanwords listed in (25)-(27) are all from the first half of the EMJ period. Many are written in hiragana in the texts, but here we also give the source kanji for reference. (25)

Lexical compounds: nama-zuryail 'governor with no real power' (nama- 'unripe, immature' + ~fijf 'governor')~ aigyail-duk- 'be charming' (~~ 'charm'+ 'attach (v.intr.)')~ nana-moji 'seven letters, characters' (nana- 'seven' + 'letter, character')~ sailzi(n)-mono 'vegetarian food'(~~ 'devotion to Buddhism~ abstention'+ mono 'thing, stuff')~ bakuti 'gambling~ gambler' (
Jt*

(26)

Early examples of native prefix or suffix with SJ lexical word: mi-kesiki 'appearance' (mi- 'beautification'+~~)~ daiji-domo 'great things' C:k* + -domo 'plural')~ taisyail-tati 'generals' (:k~~ + -tati 'plural')~ sau-gati- 'have much (grass style) cursive writing' (1it + -gati 'be frequent, likely to be/have' which derives an adjectival noun)~ rail-gawasi- 'noisy, disorderly, ill-marmered' (~l + -gawasi (< -gapasi) '-like' which derives an adjective)

9.2 Ondoku

289

Table 9.4 Naiive Japanese and SJ numerals

2 3 4 5

6 7 8 9 10 20 30 40 50 100 1000 10,000

Native

SJ

pito puta mi yo

iti ni san si go roku siti paii (> NJ hati) kiu (> NJ kyuu) zipu (> NJ zyuu)

mu nana ya kokono towo, to, -so pata miso yoso momo, -po ti yorodu

pyaku (> NJ hyaku) sen man

(27)

Adjectives sipune- 'persistent, stubborn' (fjt~); kotigoti-si- 'unrefined' (~ ~); raurau-zi- 'refined, talented' (jjjj); taidai-si'inconvenient' (:8::8:; sometimes a different etymology is suggested for this word, namely the OJ adjective tagitagi-si-); zaezae-si- 'of learned appearance' (::t::t)

(28)

Verbs saiizok- 'dress up'< sauzoku 'dress' (~JR); saisik- 'colour, paint' < saisiki 'colouring'(~-§.); saudok- 'make a fuss'< saudo(u) 'disturbance' (ftliJ), adding -k- to the SJ noun to create a verb stem; sarugaw- 'joke, jest'< sarugau < saru-gaku 'Sarugaku farce (the precursor of kyogen)' (~~)

9.2.3.2.1 Numerals The native system of numerals is simple and partly based on vowel alternations to show doubling: pito '1' ~puta '2'; mi '3' ~mu '6';yo '4' ~ya '8'. However, the system does not provide easily for formation of higher numbers. We saw above that some SJ numbers were used in OJ (9.2.3.1), but the intake of SJ numerals is usually thought not to have taken place until EMJ.

290

9 The sinification of Japanese

9.2.3.3 Sino.Japanese loanwords in Late Middle Japanese During the LMJ period the use of SJ loanwords in the texts increased. This is probably in part related to the gemes represented in the sources, including more kanji-kana majiribun, but the establishment of SJ was a major factor, making use of originally Chinese words more freely available in Japanese and thereby facilitating both intake and use of SJ loanwords. In Esopo from the end of the period (see 10.2.2), we find a sizeable proportion of SJ vocabulary, and also Vocabulario lists a large amount of SJ words. This shows that SJ words had become a well-integrated part of general vocabulary and language use by then. In addition to the increased intake and use of loanwords taken in from kan-on and go-on SJ, a new layer of J-Ch came to Japan during the first half of the period, used especially in some Zen Buddhist sects. This is the to-on variety of J-Ch (9.2.1), which also gave rise to loanwords, and also eventually to a to-on SJ. Examples of such loanwords are anzu 'apricot'~ or ~-1-; andon 'lantern' 1f1:T; isu 'chair' ~-1-Jusin 'construction'~~ (note that EMJ p > LMJ f, see 11.3). We did not above comment on correspondences between kan-on/go-on and the third, minor layer of SJ, to-on, but a look at Table 9.1 shows a stereotypical feature of to-on, the reflection of EMC *-1) as IN!, giving correspondences such as kan-onlgo-on !i, iii (> !i, u!) :: to-on IN!, as in Ji( kell kyai1 > kei/kyau > NJ kei/kyoo :: kin. As mentioned above, J-Ch to-on is said to be based on southern Chinese varieties, but the loanwords taken in during the LMJ period also include words deriving from contact between Japanese fishers and traders with their continental colleagues. In that sense, some of the words characterized as to-on are direct loans from Chinese, rather than SJ loanwords (which are based on J-Ch or SJ).

9.2.3.4 Sino.Japanese loanwords, Japano-Chinese and Sino-Japanese It is clear from the examples of EMJ SJ loanwords given above that the shapes of some SJ loanwords are not captured in the SJ kanji readings, e.g. daitoko (::kii&), pizai1 ~I=~) whose regular SJ readings are daitoku and hizyoo (< pizyai1). This reflects first of all on the non-descriptive, rationalizing nature of the established SJ readings (9.2.2.1). However it also illustrates another point, viz. that OJ and EMJ SJ loans must be thought to have been taken into Japanese from J-Ch, not from SJ. In that connection, it should also be noted that most SJ loanwords from the EMJ period which were in everyday use derive from the pre-kan-on (that is, go-on) norm of J-Ch (and therefore often correspond to the SJ go-on readings), showing the persistence of the J-Ch go-on also after the decrees promoting the use of

9.2 Ondoku

291

kan-on. Also a number of everyday SJ loanwords still in common use today, such as niku 'meat', netsu (~) 'fever', or konnichi (4- El) 'today' are based on go-on, as are the SJ numerals, except kiu (> NJ kyuu) '9' which is kan-on, but go-on ku is also used for example in counting out the numbers. Most SJ loanwords taken in during the LMJ period, on the other hand, are reading loans mainly based on SJ go-on and kan-on. The establishment of SJ made vocalization of originally Chinese words far more freely available in Japanese. This did not mean that any Chinese word could be used in Japanese, but it did mean that unfamiliar or unknown words did not sound alien and therefore they would more easily gain acceptance and currency. This also forms the background to the modernization of the Japanese lexicon during the Meiji period (17.3.2) which would not have been as smooth without the ready availability of SJ. As we saw above (9.2.3.3) a number of SJ loanwords were taken in from the new to-on variety of J-Ch which arose in the early LMJ period, and others were direct loans from Chinese. Nonetheless, since LMJ and into NJ, the major donor of SJ loanwords has not been Chinese or J-Ch, but SJ. Furthermore, many of the words originally taken in through J-Ch have since been reformed to conform to SJ pronunciation norms. Thus, daitoko and hizoo (< pizail) are no longer in use but have been replaced by daitoku and hizyoo. However, the so-called kan yo-on to some extent reflect J-Ch, in the sense that they are current SJ readings which derive from J-Ch, but which do not conform to the normative SJ readings in dictionaries. For example, tb has the following SJ readings listed in modern kanji dictionaries: to (kan-on)~ zil (go-on)~ do (kan 'yo-on). Of these the only one used in the many SJ words whose writing includes this character is do, that is, the kan yO-on. In this case the kan 'yo-on was also the reading used when new vocabulary was coined in the Meiji period (e.g. cNJjidosha 'automobile').

om

9.2.4

Sino-Japanese words arising through on-reading of a kun-writing

Some SJ words originate not in borrowing from J-Ch or SJ, or from coinage of elements borrowed from J-Ch or SJ, but rather from applying SJ readings ('on- readings') to character combinations which were originally used as a logographic representation(' kun-writing') of native words. Thus for example, NJ kazi (< kwazi) 'fire', originates in the on-reading of*-*, which was originally a kun-writing (i.e. a logographic representation), of pi-no-koto 'fire~ fire-GEN-thing'. A few examples are given in (29).

292

9 The sinification of Japanese

(29)

Earlier word

kun-writing

on-reading; SJ word

early EMJ pi-no-koto 'fire; fire-GEN-thing'

kazi (late LMJ; < kwazi)

early EMJ mono-nona(-no uta) 'acrostic poem; thingGEN-name(GEN-poem )'

butumei (ka) (late LMJ)

(OJ opo-mapye 'greatfront' >) early EMJ omape (> omawe > late EMJ omae) used as a term of address and reinterpreted as 'RESPECT-front'

go zen (early LMJ) term of address

LMJ toki-fakari (toqifacari in Vocabulario) 'clock; time-measuring'

tokei (NJ) (but cf. also tokei

early EMJ owas- and owasimas'exist.RESP'

goza (early LMJ), used in goza-ar> NJ gozar- 'be.

±~'sun-dial')

POL'

9.3

The case of:#

A single example illustrates some of the complexities involved in kanbunkundoku and ondoku and their influence on Japanese. Chinese 1!r (EMC *b;:ia') is traditionally said to be used as a pronominal head of a relative clause, 'he who ... , the fact that' (often specifically 'the sort of person who ... '), but 1!r has also been shown to be used to mark topics and conditionals (Harbsmeier

1981: 210--28). Reflecting these uses in Chinese,

1!r is in kanbun-kundoku rendered as:

(30) a. a noun particle pa 'topic'; b. the inflectional verb endings -(a)ba 'conditional' and -(e)ba 'provisional'; c. a noun mono 'person'. Conversely, in writing Japanese,

1!r was used logographically to write:

9.3 The case of ;f

293

(31) a. pa 'topic'~ b. -(a)ba conditional 'if' and -(e)ba provisional 'as, when'~ c. mono 'person' (but not mono in the meaning 'thing' which is usually written ~).

These renditions in Japanese of Chinese 1!t and the writings by 1!t of the same Japanese forms in fact provide strong support for the analysis of Chinese 1!t and more generally show how old Japanese kun-readings and -writings can reflect on and contribute to the understanding and analysis of Chinese. In writing Old Japanese phonographically, 1!t was used in the MYS as a kungana (although very rarely) for the syllable /pal (deriving from the kunreading and -writing noted above) and as an ongana for the syllable /sa/ (on the basis of its sound value in J-Ch, ultimately deriving from the EMC sound value). 1!t also at least from early LMJ gave the SJ suffix -sya which designates agent nouns '-person~ -er, -or', as reflected in for example cNJ isha 'doctor' 12S:1!t, ekisha 'fortuneteller' £1!t, kagaisha 'assailant' J.J0~1!t, gakusha 'scholar' ~1!t, kisha 'reporter' ~1!t. It is also likely that the use of mono as a pronominal relative clause head functioning as a kind of topic marker in a cleft-like construction, as in (32) from Makura no soshi, reflects influence on the use of mono from the use in Chinese of 1!t, brought about by the rendition of 1!t as mono. This construction is used in many of the beginnings of the chapters in the Makura no soshi.

(32)

susamazi-ki mono prru poyuru depressing-ACOP.ADN MONO daytime howl.ADN 'what is depressing: a dog howling in the daytime'

mu dog

Perhaps finally also some uses of the provisional verb form were influenced by the use in Chinese of 1!t, mediated by the rendition of 1!t by the provisional form, as in (33) (from the Hojoki), where oreba (or- 'be') seems to mean 'those who were' rather than 'when, as they were'.

(33)

1e oreba tatimati-ni no naka m house GEN inside DAT be.PROV instantly p1s1ge-na-mu to su be. crushed-PERF -CONJ COMP do 'those who were indoors would be crushed instantly'

294

9 The sinification of Japanese

REFERENCES Kanbun-kundoku: King and Whitman forthcoming, Kono 1969, Tsukishima 1987, Yoshida et al. 2001. Influence on Japanese from Chinese: Miller 245--6; Yamada 1935. Ondoku: Tsukishima 1993. Kan-on decrees: Yuzawa 46ff. SJ readings: Kamei 1954, Nakada and Hayashi 2000: 331ff., Numoto Tsukishima 1977: 23ff. OnEMC *-p reflected as -tu: Komatsu 1956, Unger Vance 1986. SJ loanwords: in OJ: Vovin 2005: 59ff.; in EMJ: Vovin 2003: Tsukishima 1969: 588ff, Yamada 1913/1952: 588ff.

1977, 1967: 1996: 1986, 1988, 25ff.,

Part III

Late Middle Japanese

10

Sources

The vast majority of sources in Japanese script from this period are written in kanji-kana majiribun (cf. 6.1.1). At the end of the period a number of texts in or about Japanese written in alphabet writing appear, produced by Christian missionaries, mainly Jesuits. Table 10.1 on p. 298 is a chronological list of important sources.

10.1

Early Late Middle Japanese

The sources from the first half of the LMJ period are in the main written in the increasingly fossilized classical norm which arose out of the language of the end of the EMJ period. This does not mean that innovations are not at all reflected in those sources, but they are not systematically represented. The personal hiragana literature (see 6.2.1) did not continue into the LMJ period. There is, however, a body of gunki monogatari (]I[WC.~m epic 'war tales'): Heiji monogatari (-~1-tl'~m 'The tale of the Heiji war'), Hogen monogatari C*:lC~m 'The tale of the Hogen war'), Heike monogatari (.lJL *~m 'Tale of the Heike'), Genpei josuiki (iJ»:.lJLil£:!"<WC. 'An account of the Genpei wars'), Soga monogatari (it~~m 'The tale of the Soga brothers'), Masukagami (:1:~- 'The larger mirror'), Taiheiki (.~.lJLWC. 'The record of the great peace'). Two important texts in the essayistic zuihitsu style are the HOjoki (h"XWC. 'An account of my hut') by Kamo no ch6mei and Tsurezuregusa (f;iEf~1j![ 'Essays in idleness') by Yoshida Kenko. The setsuwa literature which appeared in the EMJ period flourished in particular in early LMJ, with works such as the Ujishili monogatari (91-tl'*'.il~m 'Stories gleaned at Uji'); Hobutsushil (~~~ 'A collection of treasures'); Jikkinsho (+~Jilt-') 'A treatise of ten rules'); Shasekishil (~.:5~ 'Sand and pebbles'). Often in the setsuwa literature, dialogue parts reflect (features of) the contemporary vernacular not reflected in the frame narrative. This is a pattern which is also found in the popular literature from the Edo period. Other important materials from early LMJ include the Sangoshiiki-chil c.=.~m1~ 8:), a commentary on KUkai's Sangoshiiki (.=.~m1~), written in kanji-kana 297

298

10 LMJ sources Table 10.1 Important LMJ sources Thirteenth century Hojoki (1212) Ujishiii monogatari (1190-1242) Sangoshiiki-chii (c. 1220) Gukansho (1220) Heiji monogatari (early thirteenth century) Hogen monogatari (early thirteenth century) Heike monogatari (early thirteenth century) Jikkinsho (1252) Shasekishii (1283) lza)K)i nikki (second half of thirteenth century) Fourteenth century Genpei josuiki (early fourteenth century) Tsurezuregusa (c. 1330) Jinn6 shotoki (c. 1339-43) Soga monogatari (mid fourteenth century) Masukagami (c. 1370) Taiheiki (c. 13 72) Fifteenth century Setsuy6shii (mid fifteenth century) Rongo-sho (before 1475) Shuyaku-sho (1477) Shiki-sho (1480) Irop'a (1492) Sixteenth century Moshi-sho (1516) Mogyu-sho (1534) Shiganikkai (1534) Feiqe monogatari (1593) Esopono fabulas (1593) Dictionarium Latino-Lusitanicum ac Japonicum (1595) Rakuyoshii (1598) Vocabulario da lingoa de lapam (1603-4) Arte da lingoa de Iapam (1604-8) Arte breve da lingoa Iapoa (1620) Ars grammaticae iaponicae linguae (1632) Dictionarium sive thesauri lingure Iaponicre compendium (1632)

majiribun, the Gukansho (~IfF) 'The future and the past') by Jien, the Izayoi nikki (+:A~ 13 ~'The diary of the waning moon') a travel diary from the second half of the thirteenth century by Abutsu, and the Jinno shOtoki (t$ ~JE~~ 'An account of our divine sovereigns and true royal line'), by

Kitabatake Chikafusa.

10.2 Late LMJ

10.2

299

Late Late Middle Japanese

Two sets of sources from the LMJ period are of particular importance: the shOmono from them iddle of the fifteenth to them iddle of the sixteenth century, and the Christian sources from the end of the sixteenth and the beginning of the seventeenth century. The Korean Irop'a (1492), a basic textbook of Japanese for Korean officials, is another important external source which contributes to dating of some phonological changes (see e.g. 11.6). 10.2.1

Sh6mono

From the second half of the fifteenth century, the so-called shOmono (YY~) appear. These are commentaries, or lecture notes, on Chinese classics and Buddhist scriptures. Important texts include the Rongo-sho (~~YY), ShikishO (~~eYY), Shuyaku-shO (JaJJbt-'J;), Mogyu-shO (~>.Kt-'J;), Shiganikkai (lm MJ\#!t), andMoshi-shO (:~-1-YY). In these texts we find language which has changed greatly since the end of the EMJ period and which is very different from the language reflected in the sources from the first half of the LMJ period. It is in most respects very similar to the language reflected in the Christian sources from the very end of the period, showing that most of the changes reflected in the Christian sources had taken place by them iddle of the sixteenth century. A stereotypical feature of the didactic style in the shOmono is the very frequent use of the emphatic, or perhaps rather admonishing or impressing, sentence final particle zo. 1 0.2.2

Christian sources

Around the middle of the sixteenth century the first European we know about reached Japan, namely a Portuguese cast-away who landed in 1543 on the island ofTanegashima, south ofKyushu. Soon thereafter, in 1549, Jesuit missionaries arrived in Japan, led by Francis Xavier. Between this time and 1639 when all Europeans with the exception of the Dutch were expelled from Japan - a policy of national seclusion known as sakoku (&Jl. 00) 'closed country' that lasted until the 1850s- the Jesuits carried out missionary work with great zeal. For that purpose they produced and published a large amount of material, including texts, grammars and dictionaries, especially after they brought a printing press to Japan in 1591. They acquired an impressive competence in and knowledge about the Japanese language and collaborated closely with Japanese converts. The works produced by the missionaries and their Japanese collaborators are collectively referred to as kirishitan shiryo in Japanese ('Christian materials') while books produced on their movable-type press are referred to in the history of Japanese printing as kirishitan ban

300

10 LMJ sources

('Christian imprints')~ we adopt a similar usage here and refer to the 'Christian sources'. The work of the Jesuit missionaries constitutes one of the greatest feats on record of linguistic and cultural penetration of an unknown and alien culture. Within a short span of time the missionaries learnt both the contemporary vernacular and the classical written language, and also familiarized themselves with classical literature and scholarship. They published a number of texts, both in alphabet writing and in Japanese script. Some texts were in a simple form of the classical written language, sometimes even employing an archaizing kanbun-kundoku-like style, such as the opening of the Contemptus mundi (a translation published in alphabet writing in 1596 of Thomas a Kempis' Imitatio Christi (1418)), which uses the OJ nominal form notamauaqu /notamawaku/ to introduce a saying by Jesus, in the same way that sayings by Confucius were introduced in kanbun-kundoku renditions of the Analects (si iwaku or si notamawaku, cf. 9.1. 7). Both the original spelling and a phonemic transcription are given in (1): (1)

Von arujino notamauaqu: QUI SEQITUR ME, NON AMBULAT IN TENEBRIS, SEDHABEBITLUMENVITlE. loan. 8. Vareuo xitomonoua yamigiuo yucazu: tada jumio no ficariuo mot
zyumy::J::J no life GEN to nan CO!v.IP COP

fikari light

0

ACC

0

ACC

sit::J::J follow

motu hold

'The Lord says: 'Whoever follows me will never walk in darkness but will have the light of life.' The texts written in the contemporary vernacular are valuable sources for the language at the end of the LMJ period. The missionaries also wrote and compiled grammars and dictionaries which included information about sociolinguistic and regional differences. Their descriptions and analyses of Japanese are incisive and sophisticated, and still of practical use as well as of scholarly interest. Unfortunately their work remained largely unknown in Europe until the late nineteenth century. Had that not been the case, western Japanology would most likely have progressed significantly beyond its current stage. The work on Japanese done by the missionaries was a collaborative effort which for the most part did not identify the individuals who took part in it.

10.2 Late LMJ

301

However, two figures may be singled out for mention: The first is the Portuguese Joao Rodrigues (1561-?1633/34) who came to Japan in 1577 at the age of 16 and entered the Jesuit order in 1596. He was expelled from Japan by the shogunate in 1610 and spent the rest of his life in Macao. He is responsible for two of the most important of the Christian materials, the Vocabulario da lingoa de Iapam and the Arte da lingoa de Iapam, published in Nagasaki between 1603 and 1608. The second is the Spanish Dominican Diego Collado (d. 1638) who came to Japan in 1619, but only stayed until 1622. His Ars grammaticae iaponicae linguae and Dictionarium sive thesauri linguce Iaponicce compendium were published in 1632 in Rome. To some extent they incorporated Rodrigues's work. The material is eminently accessible for study: It is available in excellent reproductions and annotated editions, much of it in recent years also in electronic form, and the grammars and dictionaries have been published in annotated translations into modern Japanese and other languages. Issues of text tradition do not arise as these are printed texts, and the material presents no particular philological difficulties. The Christian sources have been extensively studied, both by historians and by linguists. The most extensive use of the materials has been in the areas of phonology and lexicology. The contemporary phonology, including main allophonic variation, has been well and exhaustively described on the basis of these materials, and the dictionaries form an extremely important source of the contemporary lexicon, with good exemplification of usage and with word definitions in other languages (Latin and Portuguese). The Christian sources have also been studied as part of the history of the study of Japanese, and also as part of the discipline of 'missionary linguistics'. Particular attention has been paid to the descriptive approach of applying traditional Latin grammar to the description of a typologically quite different language. Although these materials have been studied intensely, they hold a wealth ofunmined information and important data. 10.2.2.1 Texts The two most valuable texts as linguistic source material are the F eiqe monogatari (The tale of the Heike), an abbreviated translation or retelling of the popular classic originally from the thirteenth century, and the Esopono fabulas (Aesop sfables), a translation into Japanese from a Latin version of Maxim us Planudes' introduction to and version of the fables- incidentally the first piece of 'western' literature to be translated into Japanese. Both are written in the contemporary vernacular. Feiqe seems to retain some older features of the language not found in Esopo, which may be the result of influence on Feiqe from the original, either unintentional or deliberate retention of classical features. Feiqe is, however, also a single text with longer narrative passages

302

10 LMJ sources

and more varied language represented, as opposed to the short fables in Esopo, and linguistic differences betweenFeiqe andEsopo can also be taken to reflect such differences. Texts of Christian doctrine and devotional literature published by the missionary press were written in a simple form of classical Japanese which presumably lent the texts more authority. They include Contemptus mundi (1596, cf. 10.2.2), Doctrina Christan (1600) and Giya do pekadorn (1599), an abridged translation into Japanese written in Japanese script of Fray Luis de Granada's Guia de pecadores ('Sinners' guide' from 1555) which became popular and was widely read also after the expulsion of the Christians from Japan. It is the first published text in Japanese to employ the handakuten (6.1.2.3) - after its appearance in the Rakuyoshii dictionary (10.2.2.3) the previous year- in order to spell /p/ in loanwords.

10.2.2.2 Grammars Two grammars, of which unfortunately no known copies survive, are reported to have been produced shortly after the arrival of the Jesuits: Arte da lingoa Japonesa (1551, by D. da Silva) and Grammatica da lingoaJaponesa (1564, by J. Fernandez). However, the Arte da lingoa de Iapam is a monumental grammar and manual of use, edited or directed by Rodrigues and published in Portuguese in 1604-8 in three volumes over 239 folios, covering most aspects of the language to a very high standard~ it focuses on the vernacular, but also addresses the classical written language. In terms of scope and quality it was probably not superseded until Samuel E. Martin's A reference grammar of Japanese (1975), but regrettably it was largely forgotten soon after the closure of Japan to foreigners and was only rediscovered and made an object of study in the late nineteenth century. Thus, it never had much impact on Japanology in the west. Its translation into modern Japanese by Doi Tadao published in 1955 makes the Arte easily available~ note that references toArte in this book are to Doi' s translation which is clearly indexed with reference to the original. In Japanese it is usually referred to as Nihon (dai)bunten ( 13 :;ljs:(7c)Jt,$J In 1620 a revised, abbreviated one-volume edition, Arte breve da lingoa Iapoa (Nihon shobunten 13 7-js:;J,Jt$) was published. Finally, Ars grammaticae iaponicae linguae is a fine grammar of Japanese written in Latin by Collado and published in 1632 in Rome~ it is available in a translation into English (Spear 1975). 10.2.2.3 Dictionaries The Christian missionaries published four dictionaries: Dictionarium LatinoLusitanicum ac Japonicum (1595) is a Latin-Portuguese-Japanese dictionary, which was based on Calepino's dictionary of Latin which had come to serve as a template for making dictionaries since its publication in 1502~ it is not

303

10.2 Late LMJ

known who directed the compilation of this dictionary. The Dictionarium sive thesauri lingure Iaponicre compendium (1632) is a Latin-Spanish-Japanese dictionary, compiled by, or under the direction of, Collado. The Rakuyoshii (1598; ~-#:) is a dictionary of more than 2,200 kanji, organized in three parts, by on-readings and kun-readings (in Iroha order, cf. 6.1.2.4), and character shape. It is in this dictionary that the handakuten (6.1.2.3) was first used, in order to reflect /p/ in SJ vocabulary. By far the most important of the dictionaries is the Vocabulario da lingoa de Iapam (1603-4, henceforth' Vocabulario'), a Japanese-Portuguese dictionary compiled under the direction of Rodrigues. It contains over 32,000 words in a main dictionary and a supplement. It is the most comprehensive dictionary of Japanese compiled before the end of the nineteenth century and the major source of information about premodern Japanese, in particular LMJ, vocabulary. It is a sophisticated dictionary, with both word definitions and extensive examples of usage; it includes and identifies dialectal, archaic, literary, poetic, Buddhist and vulgar words, as well as words particular to women's or children's language. Verbs are listed under their infinitive, followed by the nonpast and past in abbreviated form, as exemplified in (2), thus giving the essential lexical information about verbs needed to identify their conjugational class and to form other inflected forms. Vocabulario is available in a translation into modern Japanese (Doi et al. 1980) and is in Japanese known as the Nippo jisho ( Sliim::;=). (2)

'open' 'emerge'

aqe, uru, eta ide, zzuru, eta

(= ake, akuru, aketa) (= ide, iduru, ideta)

'write' 'call'

caqi, u, aita yobi, U, oda

(= kaki, kaku, kaita) (= yobi, yobu, yooda)

REFERENCES Shomono: Yuzawa 1929/1970. Korean materials: Hamada 1970. Christian materials: Doi et al.. 1980, Sugahara 2000, Sugimoto 1989.

11

Phonology

The Christian sources in alphabet writing provide extremely valuable information about the phonology at the end of the LMJ period, including a number of features of pronunciation which were not reflected in writing in Japanese script. In addition to the system as it may be inferred from the transcription in the texts and dictionaries, the grammars (Rodrigues's Arte and Arte breve and Collado's Ars grammaticae) have notes on pronunciation. Table 11.1 shows the sound inventory at the end of the period and includes also the transcription used in Vocabulario (the transcriptions used in the various Christian sources differ on points of detail, but they reflect the same phonological system). Unless otherwise noted, LMJ forms cited in this section are from the Christian sources, mostly Vocabulario, in the transcription used therein followed by a phonemic transcription if that is significantly different, e.g. chOzzu /tyoodu/ 'washing'. Table 11.1 shows both free moras (/CV, CyV, CwV/) and long syllables with a long vowel, which in the Christian sources were noted by a single vowel with a diacritic (/CVV, CyVV, CwVV/)~ see (11.5) about the phonemic analysis of long vowels. Table 11.1 does not include long syllables whose second mora was a consonant or a high front vowel (which formed a long diphthong with the preceding nuclear vowel), as these were noted by individual segments in the Christian sources. The inventory of bound moraic phonemes is shown in (1 )~ see in particular 11.1.2 about changes from the EMJ to the LMJ system of bound moraic segments. (1)

Bound moraic phonemes II, V (or U)/ Vowels Consonants /Q, N, tl

The sound changes which took place during the LMJ period are not many or complicated. In addition to outlining those changes, we will note features of pronunciation reflected in the Christian sources. In several cases they are the only indication of a dating of features we still find in NJ, or a concrete indication of features we posit for earlier stages of the language. 304

305

11 LMJ phonology Table 11.1 LMJ sound inventory

Short CV, CyV and Cwa syllables Ia/ a

lka/ ca

/sa/ sa

Ita! ta

Ina! na

pa

/fa/ fa

/rna/ ma

Iii i,j,y

!kif qi, qui

lsi/ xi

/til chi

!nil ni

/pi/ pi

!fit fi

!mil mi

lui u

lkul qu

I 'SUI

11,

CU,

su

ltul t9U

!nul nu

pu

lful fu

lmul mu

lei ye

Ike/

lsel xe

/tel te

lnel ne

pe

lfel fe

!mel me

/so/ so

Ito/ to

/no/ no

po

Ifof fo

/mol mo

/gal ga

lza/ za

Ida/ da

lba/ ba

/gil gui

lzil ji

/di/ gi

fbi/ hi

lgul gu, gv

lzul zu

ldul zzu

lbul bu

!gel gue

lzel je

!del de

!bel be

/go/ go

lzol zo

Idol

lbol bo

/kya/ qia

/sya/ xa

ltya! cha

lnyaf nha

/fya/ fia

lmyaf mia, mea

ltyol cho

In yo/ nho

/of vo, uo

qe, que lkol co

do

!pal

lpul !pel

!pol

/pya/ pia

lya! ya

Ira! ra lril ri

lyul yu

/ru/

ru Ire! 111

/yo/

fro/

yo

ro

lryaf ria, 111a

lsyul xu

/gya/ guia

/kyo/ qio, qeo

/syo/ xo

lzya/ ja

/dya/ gia

lzyul ju lgyol guio

lzyol jo

lkwa/ qua /gwa/ gua

ldyol gio

lbya/ bia

lfyol flo

lryol rio, 1110

lwa/ va, ua

306

11 LMJ phonology

Table 11.1 (cont.)

Long CVV, CyVV and kwoo syllables /kuu/ cu, cii

sii,sii

/tuu/ t9u, t9ii

/nuu/ /puu/ /fuu/ nu, nii pii,pii fo,fii

/muu/ mu

lyuu/ yu,ya

/koo/ co

!so of so

/too/ to

/noo/ no

!pool p6

!fool .f6

/moo/ mo

/yoo/

/roo/

vO, uO

y6

r6

/oo/ vo, uo

/koo/ co

ls:Jol

/tool to

/noo/ no

/pool

!fool

/roo/

pO

fo

/moo/ mo

/yoo/

sO

yO

rO

/guu/ gu, ga

/zuu/

/duu/

ZU

ZZU

/goo/ go

/zoo/ zD

/doo/ do

b6

/goo/ go

/zoo/ zO

/doo/ do

/boo/ bO

/kyuu/ qiu, qiii

/syuu/ xU,

/tyuu/ chu, chii

/nyuu/ nhu, nhu niu, niii

/fyuu/ flu

xO

/tyoo/ cho, teo

/nyoo/ nh6, neo

/pyoo/ /fyoo/ peo, feo,flo pio

/myoo/ meo, mio

reO,

/kyoo/ qio, qeo

/syoo/ xO

/tyoo/ chO

/pyoo/ /fyoo/ pio flo, feo

lmyool mio, meo

/ryoo/ rio, reo

/gyuu/ guiu, guiii

/zyuu/ ju,ju

/dyuu/ giu, giu

lbyuu/ hiu

/gyoo/ gueo, guio

/zyoo/ jo

/dyoo/ gio

lbyoo/ heo, hio

/gyoo/ guio, gueo

/zyoo/ jo

/dyoo/ gio

/byoo/ hio, heo

fool

/kyoo/ qeo, qio

/suu/

xii,

xiu /syoo/

/ruu/ rU

/boo/

/ryuu/ riu, riii /ryoo/ rio

/kwoo/ quo

11.1

Nasality and medial voicing

In the course of LMJ changes took place to two interrelated phonetic features: medial voicing of non-initial tenues and prenasalization of the mediae were both lost. These two allophonic phonetic features were prominent elements in the sound texture of OJ and EMJ (2.2.2, 7.1.2.4). They are no longer found in

11.1 Nasality and medial voicing

307

most varieties ofNJ, but are retained in some dialects, and a reflex of prenasalization is found in the nasal allophone [IJ] of intervocalic /g/ in NJ which until recently was a regular phonetic feature, but today a normative feature of elegant diction, though not spontaneously productive for most speakers. There is no positive evidence form edial voicing in the Christian sources and judging from the fact that it is not reflected in transcription or mentioned in the notes to pronunciation in the grammars, it seems that medial voicing was lost from the language by the end of the LMJ period. It also appears that the loss of prenasalization of mediae was well under way by the end of the LMJ period. Arte and Ars gr. note that vowels are nasalized before /d, g/ and sometimes /b/ (Arte, p. 637) and /b, g/ (Ars gr., p. 5); this reflects prenasalization of these phonemes, perceived mainly as nasalization of the preceding vowel, but also indicates that prenasalization was in the process of being lost. It was through the loss of medial voicing and eventually early in NJ of prenasalization and nasality harmony that the sound texture of Japanese changed from its premodern to its modem form. This is well illustrated by (2), repeated from 2.2.2 above, which shows how different are the OJ, EMJ, LMJ versus NJ phonetics of tanabata which did not change phonemically between OJ and NJ. (2) OJ NJ

/tanabata/ [tana"bada] [thanabatha]

Loss of medial voicing and of prenasalization could well be said to be the two major phonetic changes in the history of Japanese. Apart from the direct effect on the sound texture of Japanese, loss of medial voicing was closely related to two phonological changes concerning nasality: the loss of postnasal neutralization (11.1.1) and the ensuing changes regarding phonemic nasality among bound moraic phonemes (11.1.2). Recall also that before it was lost, medial voicing was also directly involved in both onbin (7.1.4) and in the change of intervocalic /p/ to /w/ (7.3.1.1). Finally, it is possible that the loss of medial voicing was accompanied by the introduction of aspiration oftenues, which must have been an important factor in the change of /p/ to /f/ (11.3) and possibly also in the affrication of It/ before li, u/ (11.6).

11.1.1

Postnasal neutralization

The rule of postnasal neutralization (7.1.2.2) ceased to apply as an automatic phonological rule during LMJ. This is shown clearly by the occurrence of IN/ before all consonants, giving minimal pairs such as anxin /ansin/ *~ 'being safe' versus anjin /anzin/ 'spiritual peace'. Anjin is a Buddhist term,

*'i'."

308

11 LMJ phonology

which must have been lexicalized and univerbated (i.e. /anzin/ rather than /an/+/sin/), before the loss of postnasal neutralization, whereas /ansin/ '1i:. Jf 'being safe' must have been lexicalized and univerbated after the loss of postnasal neutralization. This is also the case for 'ti.'L' used in the general sense of 'relief' in the shape /ansin/, also reflecting lexicalization after the loss of postnasal neutralization. It is likely that it was the loss of medial voicing of tenues which resulted in the loss of postnasal neutralization. Although lost as an automatic phonological rule, postnasal neutralization was reflected both in morphophonological rules applying in the formation of some inflectional forms (see 12.3.2 below) and in many words which were lexicalized while the rule was still active. In addition to the few forms mentioned in 7.1.2.2, some examples are given in (3). In (3b) are shown a few examples which have variant shapes, reflecting early and late lexicalization, or later (reading) reformations, like 'ti.'L' /anzin/ versus /ansin/. (3) a.

andon D'J:T 'candle' < an + ton caronzuru 'takes lightly' < karon + suru conjiqi /konziki/ 1ft:-§. 'golden colour' < kon + siki xojo /sy-:J-:Jzyo/ 'place of birth' < syai1 + syo sanbo /sanboo/ ~~'three treasures'< san+ pou nanbocu /nanboku/ f-lij~t 'south and north'< nan+ poku.

1:m

b.

vonju lonzyu! ~m 'drinking as a (Buddhist) sin'< on+ syu; the shape vonxu /onsyu/ was also used, reflecting later lexicalization; contrast NJ inshu 'drinking (in general)' sendacu 'washing'< sen+ taku; today reformed to sentaku dojucu ldoozyuku! IP.J(E:j 'young men serving priests in temples' < doi1 + syuku; contrast NJ doosyuku 'cohabitation'

mrl

ninde /ninden/ AJ(. 'human and heavenly beings' < nin + ten; contrast zinten (which is kan-on)

11.1.1.1 'Vmu no xita nigoru' There is clear evidence, in the form of an explicit pronunciation rule, that postnasal neutralization and its lexicalized effects were part of metalinguistic consciousness. The rule is mentioned in at least two shomono, Nindengenmokusho (AJ(.BJHJF), c. 1471-3) and Gyokujin (~~, 1563), as a rule for reading out, but it is also clearly explained by Rodrigues inArte: in a section on the 'use of "sumi" [tenues] and "nigori" [mediae]', Rodrigues gives a rule whereby tenues are pronounced as mediae after 'o, 8, ii, n', that is to say, after /'J'J, oo, uu, N/, giving examples such as xoji /sy'J'JZil 'living' (< syai1 (1:.)+ si 'do.INF'), saying that 'ji' /zi/ originally is 'xi' /si/. After intro-

309

11.1 Nasality and medial voicing

ducing the rule (and noting that it has exceptions) Rodrigues goes on to say 'ao que elles dizem, Vmu no xita nigoru' (Arte, p. 633), 'which they [the Japanese] refer to as "Vmu no xita nigoru" [pronounce as a media below, i.e. after, Vmu ]'. As Rodrigues explains, ' Vmu' refers to the kana letter for u ( 5 ), which is used in the spelling in Japanese script of the long vowels o, 6, u (j':J':J, oo, uu/), and the kana letter for mu, which is often used for n (IN!) (cf. 6.1.2.5.1). Thus, Rodrigues does not describe his own observation, but reports a named rule, vmu no xita nigoru. Note that u was noticeably nasalized [ii] before m (as for example explicitly described inArte, p. 639) and that accordingly Vmu may be a bit more subtle than immediately apparent, referring, or having originally referred, specifically to [ii] (at some earlier point to Iii/) and not simply to the kana letter for u.

11.1.2

Redistribution of phonemic nasality

Among the moraic phonemes, the distribution of nasality changed from the EMJ system, where nasality was distinctive in morpheme final position for both consonants and vowels, but not distinctive in morpheme nonfinal position (7.1.2), to a system where nasality was distinctive in the moraic consonants, but not in the vowels, regardless of position in the word, see (4), that is much like NJ. In NJ word final IN/ is pronounced as an unreleased uvular or velar nasal, [N] or [IJ], but in LMJ it was an alveo-dental [n], as shown by renjo (11.4.2). (4)

/Q, N, I, V (or U)/

There were two parts to this change: First, nasality was lost in morpheme final moraic vowels, merging !0, I/ with /U, I/. In the native vocabulary, nasal moraic vowels occurred almost exclusively in the onbin-stem of verbs (see 7.1.2.2), (5a) below, and overall most nasal moraic vowels were found in SJ vocabulary, as reflexes of Chinese *-IJ (cf. 9.2.2.3), see (5b).

!0, I/ > /U, I/

(5) a.

ko'i- 'row' f. koi- 'thresh, strip off' > koitoii- 'jump' f. tou- 'ask' > tou-

b.

ke'i (Jj{) f. kei C*) koii (0) f. kou (0)

> kei >kou

Second, in morpheme nonfinal position nasality became phonemic in the moraic consonants. This second part of the change only affected the phonemic underlying representation and had little immediate effect as the phonetic

31 0

11 LMJ phonology

nasality of individual fonns remained as before. Until that change nasality in morpheme nonfinal moraic consonants had been assigned by the phonetic nasality of the following consonant, for example a prenasalized media: /akiCdo/ => [akicndo] => [akindo] (cf. 7.1.2.1 ), but as part of this change the phonetic nasality was phonemically redistributed and assigned to the moraic consonant: /akiNdo/. Conversely, in fonns such as /taCto-/ 'precious', the moraic consonant was reinterpreted as distinctively oral: /taQto-1. (6)

/C/ > /akiCdo/ > /taCto-1 >

IN//_ {b, d, g, z, m, n} IQ/ I_ {p, t, k, s} /akiNdo/ /taQto-1

In the Christian sources, /Q/ is transcribed by doubling the following consonant, e.g. maccura lmaQkural 'very black', maxxiro /maQsiro/ 'very white', etc. (or before the letter 'q' by 'c', e.g. facqua /faQkwa/ 81E 'white flower'); IN/ was mostly transcribed by 'n', sometimes before labials by 'm ', and sometimes by a tilde over the preceding vowel, e.g. ninde /ninden/ 'human and heavenly beings'. II/(< EMJ 11, II) was transcribed by 'i, j'. lUI(< EMJ /U, U/) was involved in a set of changes which produced the long vowels in Table 11.1 above, see 11.5, and was transcribed as part of those long vowels.

11.2

Loss of /w/

Loss of /w/ in the main took place in EMJ, where syllable initial /w/ was lost before /o/ and before li, e/ in word internal position (7.3.2.3). The loss of remaining /w/ before li, e/, that is, mainly in word initial position (but including also wordmediallw/ in newly lexicalized compounds), seems to have been completed during the first part of the LMJ period, probably around 1300. The change was manifested as mergers of /.wi/ f. /.if as /.if, and of /.we/ f. I.e/ as /.e/ which as in EMJ was phonetically realized as [ie] (cf. 11.8). This change also includes loss of /w/ after /k/, and the merger of /kwe/ f. Ike/ as Ike/. However, the mora /kwe/ was in any case marginal and probably not well integrated in EMJ; apart from the single EMJ word kwe- 'kick' it was found only in some SJ vocabulary. (7)

/w/ > 0 I_ {i, e} EMJ wi- 'sit down; exist'

f.

i- 'shoot'>

we 'picture' f. e 'inlet' > kwe- 'kick' f. ke 'hair'>

LMJ i Iii ye/e/[ie] qe Ike/

11.3 Fricativization of /p/; /p/ > /f/

311

Loss of /w/ in EMJ and early LMJ left /w/ only before /a/, both in syllable initial position, /.wa/, and after lk-, g-1, /kwa, gwa/, which although restricted to SJ vocabulary were more stable and widespread than /kwe/. Both lkwa, gwal are clearly reflected in the Christian sources, e.g. facqua /faQkwa/ B1E 'white flower', quaxi /kwasil ~-1- 'fruit (for dessert)', and were not lost until NJ; they are still preserved in some dialects (see 14.4). Furthermore, as part of the change of /au/> kd (see 11.5), syllable initial /.w/ was lost before/:);)/ /k;);)/ f. lkwau/ > /k.W;);)/ remained distinct.) (8)

/.w/> 0/ _ {;);)} EMJ wau 'king' f. au 'meets'>

11.3

LMJ

vo /;);)/ [w;):]

Fricativization of /p/; /p/ > If/

Through the history of Japanese, two sets of changes have affected OJ !pl. The first one, intervocalic /-p-/ > 1-w-1, took place in the second half of the tenth century (see 7.3.1.1). The second one, /p/ >Iff> !hi, started in LMJ. At some point during the LMJ period, /p/ changed to Iff, except in the position after IN, Q/ (cf. 9). The dating of this sound change is difficult for it resulted in no mergers and was not reflected in Japanese script: the kana which had been used for /pVI (t'J:t}J,...r-..f~) simply came to be used for /fV/, as well as for /p VI and for /b VI (cf. 6.1.3). (9)

/p/ >

/p/ I {N, Q} _ Iff elsewhere

Some time during NJ, Iff became !hi which is the main NJ reflex of OJ /p/ in initial position. Also that change is difficult to date, because it, too, found no orthographic expression in Japanese script (the kana f'J:l.}b.r-..1~ then, as today, being used for /hV/). However, the transcriptions in the Christian sources make clear that /p/ > Iff was complete by the end of the period, but that the stage !hi had not yet been reached. (10) shows some forms from Vocabulario: (10)

fa 'leaf'

fi 'day' June 'boat' fe 'passing' foka 'outside, beside'

(OJ pa; NJ ha) (OJ pi; NJ hi) (OJ pune; NJ hune) (OJ pe; NJ he) (OJ poka; NJ hoka).

312

11 LMJ phonology

The precise sound value of /f/ at the end of the LMJ period is not easy to establish. Remarks in Collado's Ars gr. (p. 4) indicate that it was not in all varieties of Japanese a straightforward [f], but perhaps rather [cp], possibly on the way towards [h]. (11)

The letter fin some provinces of Japan is pronounced just as in Latin, but in others as if it were approximately h- somewhere between/and h, by folding and closing the mouth and lips, but not completely. 1

This change introduced the phoneme /f/ into the phonological system of Japanese and resulted in severe restrictions on the occurrence of /p/ in the lexicon. The change of /p/ > /f/ is often said to have applied to word initial /p-1, complementing the change of intervocalic 1-p-1 > 1-w-1, but that does not take account of the time lag between the two changes. The change 1-p-1 > 1-w-1 took place in the second half of the tenth century, leaving /p/ only in initial position and after /Q/, but between that time and the change of /p/ > /f/ during the LMJ period, words with non-initial /p/ came into being, particularly through lexicalization of compounds, and non-initial /p/ in those words was also affected by the change. Thus, Vocabulario (andNJ) has many words with internal /f/ /f/, but they may all be thought to have been firmly lexicalized by the end of LMJ. (12)

yedafa /edafa/ 'branches and leaves~ digression' (yeda /eda/ 'branch'+ fa 'leaf'), safodo /safodo/ 'so, that' (sa 'that'+ fodo 'extent'), mofaya /m:>:>faya/ (NJ mohaya) 'already' (m:>:> 'already' (?< mau < ima wa 'now TOP') + faya 'early')

Three well-known words with internal LMJ /f/ > NJ /hi are afiru (NJ ahiru) 'duck' which is first attested late LMJ, afure- (NJ ahure- > cNJ afore-) 'overflow' which was OJ abure- but appears to have changed to *apure- after the change of /-p-/ > 1-w-1, but before /p/ > /f/ by which it became afore-, and (early LMJ apau >) late LMJ aj:J:J (NJ ahoo > cNJ aho) 'fool(ish)', first attested early in the thirteenth century and of unknown etymology although it looks like a SJ word. An interesting example is asafi 'morning sun' (NJ asahi) which is attested as OJ asapi, but must have been a transparent compound to escape the change of 1-p-1 > 1-w-1. Also SJ vocabulary which retained noninitial /p/ because of transparent constituency or morphemic independence (cf. 7.3.1.2) took part in /p/ > /f/. Thus- whatever the history of individual 1

I am grateful to Stephanie West for the translation from the Latin.

11.3 Fricativization of /p/; /p/ > /f/

313

words- /p/ > /f/ was not constrained to word initial position, but applied regularly to /p/ which did not occur after IN, Q/ (with a few non-phonologically defined exceptions discussed in 11.3.4).

11. 3.1

Retention of /pi after /Q/

Just like the change of internal /-p-/ > 1-w-1 in EMJ did not apply to 1-p-1 after /Q/ (7.3.1.2), the change of /p/ > /f/ did not apply to /p/ after /Q/, thus leaving -pp- (= /Qp/) sequences in addition to appare 'splendid' and moppara 'entirely' mentioned in (7.3.1.2), for example yappari (not in Vocabulario) /yaQparil 'as expected, sure enough' (thought to be related to the root yapa- 'soft, pliant' which also has reflexes in yawa-; yappari today has the variant yahari which is a hypercorrect back formation, probably based on the spelling ~~"i IJ), yoppodo /yoQpodo/ 'very, much'((?< yopodo) < yoki podo 'good extent'; is today used mainly in the shape yohodo which like yahari is a hypercorrect back formation), or cappato /kappa-to/ 'suddenly'. By far the most occurrences of /Qp/ are found in SJ vocabulary where retention of /p/ after /Q/ has been analogically extended so that all SJ morphs with initial NJ !h-1 (< LMJ /f/ < EMJ /p/) have automatic allomorphs in 1-p-1, if they are used after /Q/. Similar lexical alternations are also found, although to a lesser extent, in the native vocabulary when prefixes of the shape CVQcombine with forms with initial LMJ /f-/, e.g. (13) with intensifying maQ-: (13)

maQ-+ fadaca 'naked' fajime /fazime/ 'beginning' fira 'flat' firu 'daytime' fucura 'core'

mappadaca 'stark naked' mappajime 'origin, start' mappira 'without fail' mappiru 'middle of the day' mappucura 'core'

There are other verbal prefixes of the shape CVQ-, e.g. EMJ pilei- 'pull.INF' >piC-> LMJfiC- > NJ hiQ- after which verbs with initial /hi appear with /p/ (e.g. LMJ fippar- 'pull, drag' > NJ hippar-). All such forms may now be regarded as lexicalized and some may have been formed analogically, but the point remains that they reflect preservation of /p/ after /Q/. This is also the case with a small number of compounds which have second components with initial /pp/, alternating with initial NJ /hi, e.g. LMJ asafara 'morning-belly; stomach of someone who has not eaten breakfast' > NJ asahara - asa-ppara (not attested until the eighteenth century), the reduplicated form NJ happa 'leaf' (childish)<= ha 'leaf'(< OJ pa), or a few NJ forms which are compounds with de- 'protrude'(e.g. deppa 'buck tooth' (ha) 'tooth'), deppana 'long nose' (hana 'nose').

314

11 LMJ phonology

11.3.2

Retention of /pi after IN/

In EMJ, /p/ was neutralized as /b/ after IN/ by the postnasal neutralization rule, i.e. /INp/1 => /Nb/ (7.1.2.2), but as discussed above (11.1.1) this rule was lost in the course of LMJ. As after /Q/, /p/ was preserved after IN/ and did not change to /f/, and from LMJ onwards INp/ became a frequent sequence, as in the few examples in (14) from Vocabulario. Most words in this group are SJ, as in (14a), but a few examples are also found which include native vocabulary, see (14b): (14) a.

anpu 'safety' (*E), bonpon 'book from India' (~7-js:), monpa 'sect' (~~~).

b. binpigue /binpige/ 'hair and beard at the temples' (bin 'temple'+ pige 'beard'), bimpima /binpima/ 'right moment'- binfima (bin 1.11! 'convenient'+ pima 'time, moment') Compounds such as SJ xinforacu (f$¥*~) 'entertainment for the gods/spirits', which is one of the few words with the INf/ in Vocabulario, is a transparent productive compound from <= sin 'god/spirit' + fooraku 'entertainment'. On the whole, words with INp/ must have been lexicalized after the loss of postnasal neutralization, but before /p/ > /f/ (or have been analogically formed). Words such as those in (14) which retain /p/ after IN/ contrast with words such as sanbo and nanbocu in (3) above which have /p/ reflected as /b/ after IN/ and must have been lexicalized before the loss of postnasal neutralization.

11.3.3

Alternations arising from changes of /pi

The changes affecting /p/ have left traces in lexical alternations, see (15) which shows the main alternations within NJ: (15) a.

p- > f- > hhyaku 'hundred'

b. p- > f- > hhyaku 'hundred' hara 'stomach' c.

p- > f- > hhara 'plain'

-p(rop)-pyaku 'six hundred' -b- <= -p-

(san)-byaku 'three hundred' (biiru)-bara 'beer-belly'

(-cjl- <) -w- < -p(fuji-)wara 'wisteria plain; proper name'

11.3 Fricativization of /p/; /p/ > /f/

d.

-p- > -ppappare 'splendid! '

-w- < -paware 'evocative of deep feeling')

e.

-p- > -wuwa- 'top'

-0- < -w- < -pue 'top'

f.

-p- => -byuube 'evening' ( < yupu 'evening + pye 'side')

(-0- <) -w- < -pmae 'front'(< rna- 'eye'+ pye 'side')

315

Of these (15a) holds for all SJ morphemes in initial /hi which have (potential) variants for use after /Q/, for the few native words where /p/ is preserved after /Q/, and for forms with /p/ preserved after IN/. (b) holds for native vocabulary in rendaku and in forms with /Nb/, as well as for SJ vocabulary which lexicalized the effects of postnasal neutralization before it was lost (11.1.1). Some morphemes take part in both (a) and (b), for example hyaku ~-pyaku ~-byaku 'hundred'. (c)-(f) are all quite rare.

11. 3.4

lplin expressive vocabulary

Mimetics and other expressive adverbs seem to retain /p-1, as in many NJ mimetics, e.g. pika-pika 'sparklingly' (cf. OJ pikar-, NJ hikar- 'shine') or puka-puka 'puffingly' (cf. OJ puk-, NJ huk- 'blow (intr.)'). However, the history of mimetics and their phonemic shapes is not well described. That is especially so with regard to mimetics with /p/ because until the beginning of the seventeenth century there were no orthographic means to note /p/ to the extent this was different from /f/ after the change /p/ > /f/ had taken place (cf. 6.1.3). Vocabulario lists the following words with initial p-: (16)

pappato 'the rising of dust, waves, flames, etc.'; pararito,fararito (a) 'the sound of grain or the like falling', (b) 'all, every'; patto 'the rising and scattering of things like smoke (or spurting blood)'; paxxito (/passito/) 'for example hitting or striking the target with arrow or spear'; pinpin(to) 'the kicking and jumping of domestic animals (e.g. horses)'; pixxito (/pissito/) 'being gathered, jammed tightly'; ponpon, po(n)ponto 'the sound of beating a Japanese hand drum or letting off a rifle shot'; poppoto 'blazing up of flames etc.'.

These are all expressive adverbs, whose meanings were evidently not easy to define. The question remains, however, whether use of [p] for expressive purposes actually reflects preservation of OJ /p/, or whether these are expressive

316

11 LMJ phonology

innovations. The fact that we for some mimetics, as with pika-pika and pukupuku above, are able to identify cognate OJ verbal roots with initial /p-/ supports the former view, but the systematic productivity and elaborate sound symbolic conventions in the formation of Japanese m imetics must also be taken into account. Note for example that the equivalent in Vocabulario ofNJ pikapika is 'ficaficato' (jika-fika-to), i.e. withfrather than p, suggesting that NJ pika-pika may be an innovation. Another interesting example is 'foroforoto' in Vocabulario which, amongst other uses, is used about continuous, steady crying, as in namidagaforoforoto vochita (/namida ga foro-foro-to otita/) 'his tears fellforo:foro'; this seems to be a descendant of EMJ poro-poro, in turn probably related to the verb kobore- 'spill, flow, drop' together with which it is sometimes used, as in namida poro-poro-to kobore-tamawi-nu 'his tears droppedporo-poro' (from Genji monogatari, Yugiri). Now, NJ has both horohoro and poro-poro, in addition to boro-boro, which seem to note ascending magnitude or flow of tears, and it is not at all clear what the relation is between EMJ poro-poro, LMJ foro-foro and NJ horo-horo- poro-poro. The LMJ form suggests that regular sound change applied to EMJ poro-poro > LMJ foro-foro > NJ horo-horo; that, however, runs counter to the usual presumption of preservation of /p/ in mimetics and leaves NJ poro-poro stranded. Is NJ poro-poro then a direct reflex with retention of /p/ of EMJ poro-poro which was not noted in the Christian sources, or is it in fact a NJ innovation?

11.3.5

Phonemic split: /pi> If, pi

It would be possible to view [p] after IN, Q/ as an allophone of /f/, as [ f] did not occur in those two contexts in a (lexicalized) word, although we might find the phonetic difference between [p] and [f] too great to consider them plausible allophones of one phoneme. However, as it is, [p] was also found outside that phonological context, and we even have a rare minimal pair such as pixxi(to) (/pissito/) 'being gathered, jammed tightly' versus fixxi (/fissi/) 'pen and paper' (¥if.S:), clearly showing that a phonemic split had occurred: /p/ > /f, pl. In addition, /p/ was independently needed in loanwords (including proper names) introduced by the Christian missionaries, such as paraizo 'paradise', purutaruko 'Plutarch', or pekadoru 'sinner'. These three words are written in hiragana (l't ~ v).;e, ~ ."6 t-c.. ."6 .::. , and -""b.,~, ."6) in Giya do pekadoru (1599; cf. 10.2.2.1).

11.4

Sino-Japanese syllable final/-t/

A famous feature of the Christian sources is the use of syllable final -t in the writing of SJ words both in word final and nonfinal syllables, e.g. (17). There is little doubt that this reflects contemporary pronunciation, as also shown for example by renjo (11.4.2).

11.4 Sino-Japanese syllable final /-t/

(17)

317

fotnet suro 'get fever'(~~) connit 'today'(~ 13)

In NJ the words which in late LMJ had final /-t/ have an epenthetic vowel, usually lui (netsu 'fever'), but sometimes /if (konnichi 'today'). In earlier sources in Japanese script, these words are written with kana for tu (~)or ti ('l::, ), but it has recently been shown that a distinction also was made in some LMJ kana sources between -tu and -t, by using variant kana (hentaigana, see EMJ 6.1.2) which were originally used as equivalents for /tu/. 11.4.1

Sources offinal 1-t/

Final /-t/ is found among SJ words reflecting EMC *-t. Whereas EMC *-p and *-k were reflected in SJ vocabulary by /-pV! and /-kVI, especially in SJ readings as -pu (> -u), -ku, and -ki (see 9.2.2.3), EMC *-t was in a number of SJ words reflected as final /-t/, with no epenthetic vowel. The orthographic categories of the kana do not provide for representation of 1-t! and seem to have used kana for tu or ti for 1-t!. The use of hentaigana to represent 1-t/ is a fairly new discovery and it has yet to be seen how widespread this practice was. In NJ SJ kanji readings EMC *-tis conventionally rendered as -ti in go-on and -tu in kan-on, and it is usually thought that SJ loanwords taken in from (Japano-Chinese or Sino-Japanese) go-on had final /-ti/, whereas words originating in kan-on had /-t/, written by kana for tu and later acquiring an epenthetic /u/ to give /tu/. This is almost certainly an overly simplistic view. For example, the SJ loanword for 'fever' ~ was late LMJ net as shown by the Christian sources. In EMJ this was written as neti (cf. 9.2.3.2) whereas the NJ shape is netu. This indicates that the EMJ writing as neti was meant to represent /net/, and significantly, the initial /n-/ of these forms clearly show that they derive from go-on, corresponding to EMC *piat, cf. 9.2.2. (For~ modem kanji dictionaries give go-on as netsu/nechi and kan-on as zetu.) All this strongly suggests that this go-on based SJ loanword was net until the beginning of NJ. Another similar example is late LMJ jitni /zit-nil 'really' which was written ziti-ni in EMJ (cf. 9.2.3.2), but which is cNJjitu-ni. Again, this suggests that the form was zit from EMJ until early NJ, and also this is clearly go-on based, corresponding to EMC *~it. (For~, modem kanji dictionaries give kan yo-on asjitsu, go-on asjichi and kan-on as shitsu, showing, compared to the readings for ~, some arbitrariness in the assignment of current forms to kan yo-on or to one of the other layers of readings.) Final /-t/ is almost entirely limited to SJ vocabulary, but it is also found in a very small num her of variant shapes of native words in the Christian sources, whose regular shapes, however, have final vowels: ximot 'cane, rod (used in criminal punishment)' (- /simoto/), carafit 'big box~ lit. Chinese chest'

318

11 LMJ phonology

(- lkarafitu/); tetdai 'help' (- /tetudail). Spelling variants by hentaigana in sources in Japanese script confirm that these words had final /-t/. 11.4.2

Renjo

Final 1-t/ is also reflected in renjo, the process in which a final consonant is copied as the onset of a following syllable with initial vowel or glide (onset creation; see 7.2.1). Renjo after final /-t/ is in Arte (p. 637) said to be regular before va /wa/ which becomes ta after 1-t/, so that for example taixetta /taiset-tal is from taixetua /taiset-wal 'importance-TOP'. The only examples given inArte involve the topic particle wa, but in both shomono and the early NJ kyogen texts (13.3) there are also frequent examples of renjo after 1-t/ before the accusative particle a, which becomes to. Through the period several lexical items are also found, for example /settyoosyuu/ (SetchOshil) which is an alternative name for the mid fifteenth-century dictionary today known as Setsuyoshil /setuyoosyuu/, showing that it must earlier have been /set.yousyuu/, giving /setuyoosyuu/ by vowel epenthesis and /settyoosyuu/ by renjo). However, they are not many; the only two lexical forms included in Vocabulario with renjo after final /-t/ are xecchin /settin/ 'toilet; "tight spot"' (~~) <= setin (mentioned also in 7.2.1 and still used in NJ) and xojenbattacu /sy:J:Jzenbattaku/ 'reward good, punish evil'<= syauzen-bat.aku (lt~Fl~) which is no longer used. By contrast, renjo after IN/ appears to have been far more widespread and regular. Thus, Arle (pp. 636-7) sets it out as a general rule of pronunciation, saying that for example sanya /san.ya/ 'mountains and fields' should be pronounced sannha /san.nyal; other examples are xinnho /sin.nyoo/ 'trust' <= xinyo /sin.yoo/, ninguenna /ningen-nal 'person-TOP' <= ninguenua /ningenwa/. Also Vocabulario includes a number of forms with renjo, e.g. annan 'peace and quiet' C*~i, an + on) or innen 'destiny' (IE!~, in + en), but sometimes notes that this is the actual pronunciation, although they are written differently, or 'correctly' should be pronounced without renjo. Both Arle and Vocabulario demonstrate awareness of renjo as a general, productive rule after IN/ and of the relationship between underlying and surface form. In addition, Vocabulario, while noting actual pronunciation, recommends avoiding such forms in writing. In Esopo topic and accusative particles are always written ua, va and vo, uo, also after 1-t/ and IN/, i.e. without renjo, e.g. xujinuo /syuzin-o/ 'master-Ace' or funbetuo /funbet-o/ 'judgement-Ace', although these words would have been pronounced as from /syuzinno/ and /funbetto/. Renjo thus reflects final /-t/ and also shows that word final IN/ phonetically was [n] (as opposed to NJ where word final IN/ is an unreleased uvular [N] or velar or [IJ]). The difference in application of renjo after IN/ and /-t/ also indicates a difference between those two phonemes in the language, suggesting that final /-t/, which was (almost) only found in SJ vocabulary, was not as

11.5 Long vowels

319

stable as other syllable final phonemes, and that it was in the process of acquiring an epenthetic vowel.

11.5

Long vowels

Through the LMJ period, long diphthongs of the shape /Vu/ changed as shown in (18a), which also gives the further developments in NJ. These changes are in Japanese known as choon-ka (ffi:if-1~ 'change to long sound'), exemplified in (18b), which gives the spelling in Vocabulario followed by a phonemic notation; see further below about the phonemic interpretation of the long vowels. (18) a.

(late) EMJ liu/

>

leu/ /au/ foul

LMJ /yuu/ /yoo/ /';)';)/

>

NJ /yuu/ /yoo/

fool

fool fool

LMJ yu lyuu/ urexu /uresyuu/ aqiudo /akyuudo/

NJ yuu uresyuu akyuudo

eu 'gets drunk' keu 'today' teiidu 'wash(ing)'

yo /yoo/ qeo, qio Jkyool chOzzu /tyoodu/

yoo kyoo tyoozu

auti 'Japanese bead tree'

ooti

wail 'king, emperor'

vochi h':Jti! [w ';) :tSi] voh':JI [w':J:]

00

(SJ ::£.) yaii 'way of doing'

yo IY"";

yoo

taUto- 'precious'

toto- lt':J':Jto-1

too to-

ougo 'protection'

vogo loogol [wo:go] voso /osoo/ cogi Jkoodi/

00

b. Examples (late) EMJ iu 'says' uresiu 'long ago' akiiido 'trader'

(SJ~)

osou 'slowly, late' koiidi 'lane, alley'

osoo koozi

These changes can be said to have introduced long vowels in the phonological system, although it must be thought that Iii/ and /uU/ (which arose in EMJ both through onbin (7.1.4) and through loss of /-p-/ before lui and of /-p-, -w-1

320

11 LMJ phonology

before Iii (7.3.1.1, 7.3.2.3)) were pronounced as long vowels all along. We assume here that the changes to long vowels took place after the loss of distinctive nasality in the moraic vowels (see 11.1.2). There is, however, no evidence either way and both orders are possible. The difference between the LMJ and NJ long vowels is that EMJ /aU/ and /oU/ gave two distinct long vowels in LMJ, which later merged as NJ fool. They are consistently transcribed differently in the Christian sources, as (EMJ /aU/ >) o and o (< EMJ /oU/). The difference in vowel quality is usually thought to have been (/aU/>) [:>:]versus [o:] (< /oU/). Both were rounded vowels and had a phonetic onglide in syllable initial position, as shown by the transcriptions v6, ui5 [w:>:] and vo, uo and [wo:] (like short syllable initial /.o/, transcribed by vo, uo), see below (11. 8)).

11. 5.1

Phonemic analysis

In (18) and in Table 11.1 above the long vowels are transcribed by double vowels. More abstractly long vowels in Japanese are often analysed phonemically as consisting of a nuclear vowel followed by an 'empty' moraic vowel or length phoneme, noted variously as N !, IR/, or !HI, e.g. kyoo 'today' !kyoVI, kyoR/, or !kyoH/. A problem in applying that analysis directly to the long vowels in LMJ is that i5 [:>:] then would be /:>VI (or /:>R, :>HI), but that would be structurally unusual as the short vowel hi was not a part of the system. It is alternatively possible to view o [:>:] as the realization of a long a, i.e. as being a realization of /aVI (or /aR, aH/). Support for that may be found in forms such as the exclamatory final particle no In:>:>/ which is usually thought to be from a lengthened version of na, or the mesial demonstrative adverb so /s:>:>/ which comes from sa (see 12.5), but on the other hand these are singular developments, possibly analogically motivated. Against the analysis of o [:>:] as /aVI speaks the fact that o [:>:] clearly was a rounded vowel, with a phonetic onglide when in syllable initial position (see 11.8). It is also possible to view o [:>:]as a phonetic realization of underlying /aU/. This has the advantage of providing a source in the phonemic representation of the phonetic rounding. It may find support in the fact that the changes to long vowels also gave rise to morphophonological rules for verb and adjective forms ending in -u, whereby for example the conclusive of 1-w/ base verbs was realized with a long vowel: aw-u 'meets' => vo /:>:>/, ow-u 'carries' => vo fool (see further 12.3.3). On that analysis, the changes to long vowels reflected in the Christian sources at the end of the LMJ period were mainly changes in the phonetic realization of diphthongs whose outcomes were later phonemicized as reflected in NJ. In any case, o [:>:] seems to have been fairly short-lived in the language and merged with [o:] in the first part of the NJ period. In (19) we show two different phonemic analyses of the long vowels:

11.5 Long vowels

321

Table 11.2 Sources of S.Jkoo and yoo EMJ EMC

;;c "fl

w )I[; 0 j;b

1}

~

.ffl ~

~

*kaiw/kz:w *kaip/kz:p *xiaJ;J *kwaJ;J *k~w'

*kiap *k:IWJ;J *jiaJ;Jh *juawJ;Jh *'i'jiaw *jiap

(19)

EMJ

LMJ

-p->0/_u

denasalization

kau kapu

kau kwail kou kopu kou yail you eu epu

>

kau kau kwau kou kou yau you eu

LMJ

>

NJ

monophthongization

();) > 00

koo koo koo kwoo koo koo koo

koo koo koo koo koo koo koo

yoo yoo yoo yoo

){}0 ){}0

)UO )UO

NJ

liul [iu] [Ciiu]

lyuul [yu:] [Cyu:]

lyuul [yu:] [cyu:]

leu! [ieu] [Cieu]

!you! or /yoo/ [yo:] [Cyo:]

lyool [yo:] [Cyo:]

/au/ [au] [Cau]

/au/ or !':J':J! [w':J:] [C':J:]

fool [o:] [Co:]

foul [wou] [Cou]

foul or fool [wo:] [Co:]

fool [o:] [Co:]

Regardless of the phonemic analysis of the changes in (18) and their outcomes, they consist of a decrease in contrast in long /VU/ diphthongs between the peak vowel and the moraic vowel following it and may for that reason be characterized as monophthongizations.

11.5.2

Sources of long vowels in Modem Japanese Sino-Japanese

Before the monophthongizations, a sizeable proportion of SJ vocabulary had !Vul diphthongs, themselves from a number of sources, and the monophthongizations therefore resulted in widespread merger within SJ vocabulary of

322

11 LMJ phonology

previously distinct syllables. Table 11.2 shows the sources of SJ koo and yoo. Changes preceding the monophthongizations include loss of 1-p-1 before lui (EMJ) and loss of nasality in 1-fil (LMJ).

11.6

Assibilation and palatalization

Through the LMJ period, It, d/ became phonetically assibilated before the high vowels, li, u/, giving allophonic rules as in (20), which also shows the transcription in the Christian sources. (20)

It/=> /d/ =>

t~

WI I- u [tS] I_ i

ltu/

[dz]/_u [d 3] I_ i

/du/ zzu /di/ gi

/ti/ chi

The onset of the phonetic assibilation is difficult to date precisely as it was not reflected in writing in Japanese before the Christian sources, but the Korean Irop'a from 1492 transcribes Japanese /ti, tu/ by "CJ /ti/ and T- /tu/, showing that assibilation had not yet set in by then. The phonetic assibilation of /d/ eventually resulted in the merger of /d/ and /z/ before li, u/, but that phonemic change is not thought to have been completed until NJ (see 14.1), as the Christian sources keep /d/ and /z/ distinct before li, u!, noting /zi/ as ji, /di/ as gi, /zu/ as zu, and /du/ as zzu.

11. 6.1

Palatalization

Consonants remained palatalized before the front vowels li, e/, as in EMJ (21): (21)

C => [Ci] I_ {i, e}

It is directly reflected in the transcriptions used in the Christian sources (see Table 11.1) that /s, z, t, d/ were palatalized before /i/ and /s, zl before /e/: /si! xi, /zi/ ji, /til chi, /di/ gi, /se/ xe, /ze/ je. Palatalization before li, e/ is not as clearly and directly reflected in the transcription of other short syllables, e.g. short /mi/ mi, !nil ni, /me/ me, /ne/ ne. However, on closer inspection the facts point to palatalization of all consonants before /i, e/: Short syllables with palatal glide, /OjV/, are transcribed by CiV or in some cases alternatively by CeV, e.g. /kyo/ qio, qeo, /mya/ mia, mea, /ryal ria, rea, showing that both Ci- and Ce- were used to represent /Cy-/ and suggesting that the consonant was palatalized in Ci and Ce used to represent single moras, e.g. /mi/ mi ' The alternative transcriptions of /ki/by qui (rather than qi) and Ike/by que (rather than qe), which could be taken to suggest non-palatalized /k/ before /i, e/, are very rare. Note that asgi was used to transcribe /di/ [
11.8 Onglides

323

[mii], /me/ me [mieV In the transcription of long syllables /Cyoo/ or !Cy:d, equivalence between Ci- (qi-, gui-,fi-, bi-, pi-, mi-, ri-) and Ce- (qe-, gue-,fe-, be, pe-, me-, n?-) is even more widespread. Also the transcription of /nyuu/ by nhU, nhU, niu, niu and of /nyoo/ by nhO, ne6 shows equivalence between nh-, ni-, ne- to note /ny-1; likewise, /ty-/ is transcribed by both ch- and te-. Finally, also the fact that /Ciu/ and /Ceu/ changed to /Cyuu/ and /Cyoo/ (see 11.5) shows that consonants were palatalized before li, e/.

11.7

Final vowel deletion

A phonetic feature described by Collado in Ars gr. (p. 5) which is familiar from NJ is the weakening and deletion of word finalli, u/. Collado describes this as a general phenomenon and gives examples such as jitot~u 'one' pronounced as jitot~ [fitots], axi no fara 'reed plain' => ax [aS] no fara, and gozaru 'exist, polite' => gozar (which interestingly shows word final vowel deletion after /r/, which does not happen in cNJ). Affrication and palatalization of the consonant before li, u/ show that final vowel deletion was a very low level phonetic rule which took place after other allophonic variation had been assigned. It should be noted that final vowel deletion thus is not related to, or involved in, the occurrence of syllable final 1-t/, as the many forms with syllable final /-t/ did not have palatalization or affrication.

11.8

Onglides

At the end of LMJ, syllable initial I.e/ and /.o/ were still, as in EMJ, pronounced with an onglide.

(22)

I.e/=> /.of=>

This is clearly shown by the transcriptions in the Christian sources, ye and vo, uo, and also by the fact that/.eu/ > /.yoo/. Also the new long vowels [o:] and [:J:] (11.5) had labial onglides in syllable initial position, as shown by the transcriptions vo, uo and v6, u6. Functionally, the onglides served to signal syllable boundaries, creating a phonetic onset in a vowel initial syllable. This is for example shown by the distinction, clearly reflected in the transcriptions, between /o.o/ belonging to two different syllables and fool in a single long syllable (whether it is ultimately analysed phonemically as /oV/ or /oU/, see 11.5.1), as in (23):

324

11 LMJ phonology

(23)

Hetero-syllabic /o.o/

Homo-syllabic fool

touoi [towoi] /to.oi! 'distant'

toi [to:i] /too.il 'Eastern barbarian' (JI:t:~)

qiouo [kpwo] !kyo.o/ (blfl) 'falsehood + Ace'

qeo, qio 'today'

[~o:]

/kyoo/

While in general the onglides served to maintain the syllabic independence of /.of and I.e/, there are a few sporadic changes where /.o/ or /.e/ have been incorporated as a second mora into a preceding syllable, e.g. vo [wo:] fool 'big' (< EMJ o.o < owo- < opo-), toea [to:ka] /too.ka/ 'ten days' (< EMJ to.o < towo~ the word 'ten' itself is listed separately as touo [towo] /to.o/, i.e. as dissyllabic)~ or cairu lkairul 'frog'(< EMJ kaweru < kaperu~ the NJ form is kaeru and Vocabulario gives cayeru as an alternative form, but says that cairn is the colloquial form). A similar phenomenon is observed when syllable initial /a/ followed short syllables ending in /i, e/: The transition between the two syllables seems to have produced a palatal glide which became the onset in the second syllable so that li.a, e. a/ became /iya, eya/~ this was sometimes noted in writing, both in the Christian sources and in materials in Japanese script, for example miyacaxi /miyakasi/ 'votive light' (< mi-akasi 'HON-light'). Most examples are verbal compounds with the reciprocal auxiliary verb -aw- 'do together' further changed to -y::xJ (< {-il-e)-au): caqeyo /kakey~~/ 'matches, is equal to' ( < kakeau), cami-yota /kami-y~~ta/ 'bit each other'(< kami-auta). 11.8.1

Diphthongal asymmetry

The monophthongizations (11.5) are the last in a number of sound changes which occurred since OJ which have resulted in an asymmetrical distribution of diphthongs, favouring /yV, Vi/ diphthongs over /wV, Vu/. From EMJ the language acquired NUl and /VI/ syllables, with both of II, U/ occurring after all vowels (7.1.1). However, in the course of the monophthongizations in (18), monomorphemic /Vu/ disappeared from lexical representation~ it remained in morphologically complex forms, which, however, always have a morpheme boundary: /V-u/, e.g., kau 'buys' <= kaw-u (see 12.3.3). InNJ, long syllables ending in /if are often monophthongized in casual speech (and regularly in all registers in some dialects), for example in standard NJ as follows (with other outcomes in some dialects): keito 'wool'=> [keeto], nai 'non-existent'=> [ne:], hidoi 'terrible'=> [hide:], samui 'cold'=> [sami:]. It is likely that monomorphemic /ei/ has changed phonemically to /eel, so that for example the phonemic shape of sensei 'teacher' is Isensee/, but monophthongization of the other diphthongs is in most NJ dialects probably to be

11.8 Onglides

325

regarded as phonetic realizations from underlying /ai, oi, ui/. Thus, /Vu/ diphthongs disappeared within a fairly short span of time after coming into being, whereas /Vi/ diphthongs are affected by such changes at a much slower pace, if at all. Conversely, whereas /Vu/ diphthongs were eliminated in the course of the changes in (18), syllables with a post-consonantal palatal glide (/Cy-1) which through EMJ had been restricted to SJ vocabulary, gained in lexical frequency through the changes Ciu > Cyuu and Ceu > Cyoo, which introduced these syllable types also in native vocabulary. OJ had most combinations of both of the onset glides, /w, y/, and nuclear vowels, li, e, a, o, u/, the two - symmetrical - exceptions being the absence of /wu, yi/ ( cf. Table 2.4); on the phonemic interpretation of the ko-otsu distinctions given in 2.1.4, OJ also had /Cwi, Cwo, Cye/. However, since the end of OJ, /w/ has gradually been lost both in /Cw-/ and /.w-/ contexts (cf. 7.3.2, 11.2, 14.4). This has resulted in limitations on the occurrence of the labial glide /w/, such that it at the beginning ofNJ had disappeared before all vowels except /a/, whereas the palatal glide /y/ remained before all non-front vowels, /a, o, u/ (see Table 11.1). This pattern of change and distribution has to some extent been upset by very recent loanwords in contemporary NJ (e.g., wirosu 'virus', wesutan 'western (movie)', wokka 'vodka'), but also there it has been observed that /wV/ and /Vu/ are less stable than /yVI and !Vi/. REFERENCES Doi eta/. 1980, Imaizwni 1971, Mabuchi 1971, Martin 1987, Wenck 1959. Final /-t/: Morita 1993: 196ff.; final 1-t/ and hentaigana: Sugahara 2000: 21--82. Renjo: Morita 1993: 209ff. Diphthongal asymmetry: Kubozono 2008.

12

Grammar

Whereas the main phonological changes between OJ and NJ took place during the EMJ period, it was during the LMJ period that most of the significant grammatical changes took place which transformed Japanese from its premodern to its contemporary shape in both morphology and syntax. The course and precise dating of some changes is difficult to trace through the written sources; they are mainly observable in the sources dating from the end of the period. It may be no coincidence that sweeping changes took place during a period of civil war and great social upheaval and change which also would have resulted in a relaxation of social and linguistic norms.

12.1

Verb morphology

12.1.1

Inflected verb forms

Table 12.1 shows the main inflected verb forms at the end of the LMJ period. Comparing this inflectional paradigm with that ofEMJ in 8.1.1 above shows that among the finite forms from the EMJ paradigm the only surviving category is the imperative, which acquired a new variant shape in -i with the vowel base verbs, alternating with the older ending -yo; see 12.3 below for the shapes with all conjugation classes. Other EMJ forms were either lost or reinterpreted, and new forms had appeared. The loss of the exclamatory, conclusive and adnominal verb forms will be discussed in 12.1.2 and the emergence of the nonpast, past, intentional and volitional forms in 12.1.3. Other lost forms were used through most of LMJ, but were either entirely or almost lost by the end of the period, as reflected in Feiqe and Esopo: The prohibitive na . .. so was used through LMJ, but was on its way out towards the end of the period. It had by then to a large extent functionally been replaced by the prohibitive final particle na which is still in use in NJ. The negative conjectural -(a)zi, too, is rare and had largely been replaced by mai-mazii 12.1.4.1). Optative -(a)namu was lost during LMJ, whereas the newer EMJ optative -(a)baya is found sporadically even in Esopo, but on the whole it has been replaced by the new desiderative auxiliary -(i)ta- (12.1.5.3). Among the 326

12.1 Verb morphology

327

Table 12.1 LMJ inflected verb forms

Basic stem

QD

LB

kak-

alee-

kaku kaita kabazuru kakaa kaitaraa kake

akuru aketa akyoozuru akyoo aketaraa akei- akeyo

kaki kaite kakaba (- kaitewa) kakeba kakedomo - kaitemo kaitara(ba) kaitareba kaitaredomo kakaazureba kakaazuredomo

alee akete akeba (- aketewa) akureba akuredomo - aketemo aketara(ba) aketareba aketaredomo akyoozureba akyoozuredomo

Finite Nonpast Past Intentional Volitional Past conjectural Imperative

Non-finite Infinitive Gerund Conditional Provisional Concessive Past conditional Past provisional Past concessive Intentional provisional Intentional concessive

non-finite fonns, continuative -{i)tutu was no longer used productively~ its uses in cNJ, where it amongst others is used to fonn a (written) progressive on telic verbs (shinitsutsu ar- 'be dying'), seem to be revived from the classical written language. Conditional -(a)ba was still in use at this time, but was in decline. Conversely, incipient conditional uses of the provisional -(e)ba are found at the end of the period and in addition, the conditional form of the copula, naraba, was starting to be used as a conditional particle, reflected in NJ as nara. Finally, a number of new tensed fonns had appeared, see (12.1.3.3), including a new past conditional -(I)tara(ba), which eventually gave the NJ conditional -(I)tara. 1 12.1.1.1 Gerund In addition to its continued use as a subordinate verb form, the gerund came to be used in grammaticalized serial verb constructions during LMJ. In OJ and EMJ, the infinitive was used in a few such constructions, for example with mi- to fonn experiential expressions (mi miru 'tries looking'), or with positional verbs in a stative fonn to form progressives (mi witari 'is looking'~ see 1

It is sometimes alternatively suggested that the NJ conditional -(!)tara orginates in the LMJ past provisional-(J)tareba (> -(I)tarea > -(/)tara), and likewise that the NJ conditional particle nara is not from naraba, but from provisional nareba > narea > nara (Iwai 1973: 153, 181).

328

12 LMJ grammar

8.4.2.1), but with the exception that existential verbs were used with a verbal gerund to form periphrastic or analytic statives (3.1.4.7.3) and with the copula gerund nile(> de) to form extended copula forms (8.2.3), the gerund was not used much in such constructions. However, from LMJ a number of other verbs habitually came to combine with the gerund in grammaticalized serial verb constructions, for example VERB-le mi- 'try VERB-ing' (mi- 'see') and VERB-le ok- 'do and leave; do in advance' (ok- 'put') which are still used in cNJ, and benefactive constructions such as VERB-le kure- 'VERB for me/my sake' (kure'give (me)'), VERB-le kudasare- 'VERB.RESP for me/my sake' (kudasare- 'give (me) (RESP)'), VERB-te tamawar- 'id.', VERB-te tabasim- 'id.'. This type of analytic construction is widely used in NJ, with a number of different verbs. 12.1.2

Loss of exclamatory, conclusive and adnominal

The exclamatory form was already in the second half of EMJ used only as the predicate in kakari-musubi constructions with the kakari particle koso (8.9.2) -as opposed to OJ and parts of early EMJ where it was used outside kakari-musubi with koso - and the form had already then lost independent function. Although the actual shape is found through most of the LMJ period, it thus already by late EMJ did not represent an independent morphological category, but was simply a conditioned variant of the conclusive, used in agreement with koso. In the following we refer to this as the 'exclamatory variant' (of the conclusive). In the course of the loss of kakari-musubi (12.6.1.2), the exclamatory variant came not to be used at all and disappeared from the language as a word form. However, it continued to be used as a stem on which some inflected forms were built (see 12.3). In both Feiqe and Esopo, the exclamatory variant is found in agreement with koso. In Esopo, the predicates used with koso are mainly grammatical forms which belong to or diachronically reflect the r-irr conjugation (such as are 'exist', gozare 'exist. POL', nare copula, -{l)tare 'past') or the intentional -(a)uzure, but also with those morphemes there are exceptions, where the predicate is not in the exclamatory variant, but the nonpast (see 12.6.1.2.3). In Feiqe there are more cases than in Esopo of lexical verbs with an exclamatory variant in a predicate correlating with koso, but there are also a great deal of exceptions, that is, predicates correlating with koso which are not in an exclamatory variant. The use of koso in shomono texts is like that in Esopo. That suggests that the use of the exclamatory variant in Feiqe may have had archaizing features, originating perhaps from the fact that it was a translation or retelling of an early LMJ text. Kakari-musubi with koso is no longer found in the early NJ kyogen texts. The loss of the conclusive and adnominal categories took the course of a merger between the two forms (henceforth 'concl/adn merger'), in which the

12.1 Verb morphology

329

Table 12.2 Merger of EMJ conclusive and adnominal EMJ

LMJ

Conclusive

Adnominal

Non past

r-irr n-irr LB UB s-irr k-irr

ari sinu aku oku su ku

aru sinuru akuru okuru suru kuru

aru sinuru akuru okuru suru (- su) kuru

QD

kaku miru keru

kaku miru keru

kaku miru keru

UM LM

adnominal increasingly came to be used to conclude main clauses, outside kakari-musubi. Such conclusive uses of the adnominal are found in small numbers in OJ and increasingly in the first half of EMJ, where, however, the use of the adnominal in this position is said functionally to have been exclamative (cf. 8.9.4). However, through late EMJ and LMJ, the use of adnominals in conclusive position increased greatly. In the course of this the exclamative effect faded and was lost and eventually the adnominal replaced the conclusive altogether, resulting in the loss of the morphological distinction between the two forms and their functions, see Table 12.2. This may be thought to have happened in the first half of LMJ. The resulting form took over most of the syntactic functions of both the conclusive and the adnominal, e.g. as predicate in main clauses and in adnominal clauses; this will be discussed in more detail below (12.6.1). For the r-irr verbs, the conclusive was the only form which was distinct from the QD verbs, and as a result of the concl/adn merger, the r-irr verbs merged into the QD verbs. As opposed to other verb classes, the s-irr verbs used both conclusive (su) and adnominal (suru) shapes in variation through LMJ, with both variants being used in both adnominal and main clauses, but towards the end of the period suru was settled upon as the main variant. However, a reflex of the conclusive is still found as a variant at the end of the period, especially when followed by negative conjectural mazii - mai (su mazii, su mai 'probably doesn't do') and prohibitive na (su na 'don't do', which however alternated with suru na); these usages of the variant su are still found in cNJ. Particularly the EMJ intentional auxiliary -(a}fize- (8.5), which belongs to the s-irr conjugation, used both original conclusive (-uzu) and adnominal (-uzuru) shapes until the end of LMJ; from the second half of LMJ both shapes were used as

330

12 LMJ grammar

Table 12.3 Loss and changes ofEMJ auxiliaries EMJ

LMJ

-(q)zu negative

-(q)n- negative

NJ

-(i)n- perfective -(i)te- perfective -er- stative -(!)tar- stative >

-(l)ta past/perfect -(l)tara(bq) past conditional (-(l)tari representative)

-(l)ta past -(!)tara conditional -(l)tari representative

-(a)m- conjectural>

-(a)u volitional

-(y)oo volitional

-(a)iize- intentional>

-(a)uzuru intentional

-(l)tara-m- stative-conjectural

-(I)tarY.J past conjectural

-(i)ki simple past> -(i)ker- modal past>

-(l)taroo past conjectural

variants in both main and adnominal clauses. As late as in Esopo, -uzu is used in sizeable num hers, although the main variant is -uzuru, and there are tendencies in the use of the variants which seem to reflect the original distribution. 12.1. 3

Loss and change of auxiliaries

In the course of LMJ, most of the remaining auxiliaries described in 3.1.4.5 were either lost, or changed from auxiliary to flective (recall that subjunctive -(a)masi was lost already in EMJ). The only exception is the negative, which remained an auxiliary into NJ (12.1.5.1). Of the OJ auxiliaries in 3.1.4.5, most were lost entirely from the language in the course ofLMJ, leaving productive reflexes of only two of these auxiliaries which were so important to the language in OJ and EMJ: (a) conjectural -(a)m- gave the volitional ending -(a)u which is the reflex of the onbin variant of the EMJ conjectural conclusive -mu!U. In NJ -(a)u > -(y)oo (see 15.1.1). And (b) EMJ stative -(!)tar- is reflected in the past/perfect flective -(I)ta (> NJ -(I)ta), in the past conditional -(I)tara(ba) (> NJ -(I)tara conditional), and in the NJ representative -(I)tari of which there are a few examples in LMJ, but which does not become established as an inflected form until NJ. The combination of stative and conjectural, -(I)tara-m- is reflected in the LMJ past conjectural -(I)taraa (> NJ -(I)taroo past conjectural). In addition, the (late) EMJ intentional-(a)iize- (see 8.5) was reflected as an inflectional ending -(a)uzuru. The lost auxiliaries are individually reflected sporadically in some dialects, but the OJ and EMJ systems of auxiliaries are not reflected in any dialect.

12.1 Verb morphology

331

12.1. 3.1 Morphological categories The perfective auxiliaries, and the category of perfective, were lost altogether. Already in EMJ the two perfectives had partially acquired different functions and distribution (see 8.4.2), with -(I)te- being used as a recent past in some contexts; through LMJ both auxiliaries declined in use and were eventually lost. At the end of the period, -tu (the conclusive of -(I)te-) is reflected in two uses: (a) in alternative or enumerating constructions Vrtu VTtu se- 'do V 1 and V2', functionally corresponding to the NJ representative in -(I)tari (see 15.1); and (b) in the flective -(I)turaa '(recent) past conjectural' (see further 12.1.4). The EMJ auxiliaries expressing past tense were lost in the course of LMJ and the past tense function shifted to the reflex of the EMJ stative -(!)tar-. The two EMJ past tense auxiliaries, -{i)ki simple past and -{i)ker- modal past, and the use of -(I)te- as a 'recent past' auxiliary, continued in the first half of LMJ, and after the concl/adn merger the EMJ adnominal shapes of the simple past, -si, the modal past, -keru, and the recent past, -turu, were used, but they were lost before the end of the period. As mentioned, the recent past was reflected in late LMJ -(I)turaa '(recent) past conjectural'. The simple past is reflected in some dialects, but is lost from standard NJ. The modal past is reflected in the emphatic particle -kke which is used in cNJ after past -(I)ta (-(I)ta-kke) and copula da (da-kke), reflecting LMJ -takkeru (< tari-keru) and deari-keru, with drop offinal-ru (see 12.4.1). Alongside the decline of the earlier past tense markers, the reflex of the EMJ stative -(!)tar- was reinterpreted from an aspectual marker to combine the functions of past tense and perfect. The timing and course of this are difficult to establish, as is the causal relationship, if any, between the decline and loss of the EMJ past tense marker and the acquisition of past tense function of the EMJ stative. Already in EMJ, -(!)tar- was spreading at the expense of the older OJ stative -yer-, which started losing its productivity in the EMJ period, although it remained a part of the written classical language. With the loss of -yer- and the change of -(!)tar- past/perfect, the functions associated with the statives (see 3.1.4.7) came to be expressed solely by analytic constructions, mainly gerund -(I)te + existential verb (see 12.4). Identical constructions (-(i)te ar-) were the OJ origin of the EMJ auxiliary -(!)tar- (8.4.2), and clearly this is a pervasive and recurring constructional pattern in the language. The OJ and EMJ conjectural auxiliary -(a)m- gave the LMJ volitional form -(a)u. The OJ!EMJ conjectural had both volitional and conjectural uses, but these were now divorced so that -(a)u generally came to be used in volitional and future functions. However, with existential verbs and the regular and adjectival copula, the reflex of the EMJ conjectural retained conjectural function. The combination of stative (or past) and conjectural, -(I)tara-m- became

332

12 LMJ grammar

a past conjectural, -(I)tar:J:J; see further 12.1.4 about the expression of conjectural meaning. Note that even in cNJ the volitional is occasionally used in conjectural function in classical-flavoured writing. The EMJ intentional auxiliary -(a)iize- (see 8.5) >early LMJ -(a)uze- gave the late LMJ intentional form -(a)uzuru which overlapped functionally to some extent with volitional -(a)u. The intentional disappeared in the course of NJ and is no longer found in cNJ.

12.1.3.2 From auxiliary tojlective; inflectionfor tense A significant part of the changes affecting the EMJ morphological system was a shift of EMJ -(I) tar-, -(a)m-, -(a)iize- and -(I)tara-m- from auxiliary to flective: LMJ -(I)ta, -(a}u, -(a)uzuru, -(I)tar:J:J, and the concomitant change of the outcome of the concl/adn merger to a nonpast tense. The pivot in this change was the concl/adn merger. Conclusive and adnominal were the only finite categories these auxiliaries inflected for- they had no imperative and the exclamatory was at the beginning of LMJ no longer a distinct morphological category, but a variant of the conclusive (12.1.2). Thus, once the conclusive and adnominal merged, these auxiliaries no longer had distinct forms expressing distinct finite morphological categories, and they became flectives. This was, however, not simply a shift in morpheme type of these individual morphemes. It first and foremost constituted a change in the categories expressed by basic inflected forms in the system. Before the concl/adn merger both lexical verbs and auxiliaries inflected for conclusive and adnominal; when that distinction merged and the auxiliaries changed to flectives, the outcome of the concl/adn merger in the lexical verbs, the flective -u - -uru, no longer combined with the reflexes of -(I) tar-, -(a)m- and -(a)iize- to form one of their inflected forms, but instead came to form oppositions with them - it was reinterpreted as a nonpast, in opposition with past, intentional, volitional and imperative. This is shown in Table 12.4. (See further about the shapes of the LMJ past in 12.3.2 and 12.3.3, and volitional and intentional in 12.3.3.) In this way, verbs came to inflect for categories which earlier had been expressed by auxiliaries, first of all tense; thus the representation of these categories became obligatory, whereas it earlier had been optional (cf. 3.1.4).

12.1. 3. 3 Non-finite forms The new inflected tense distinctions were also represented among the nonfinite forms, see (1). This extensive system reflects the origin of the tense markers in auxiliaries and arose in the transition between auxiliarization

12.1 Verb morphology

333

Table 12.4 Changes of EMJ auxiliaries to LMJ jlectives EMJ kllk-

'write' kai-tar+STAT

kllka-m+ CONJ

kllka-iize+ INT

kai-tara-m+ STAT-CONJ

like-

'open' like-tar+STAT

llke-m+ CONJ

llke-iize+ INT

llke-tara-m+ STAT-CONJ

LMJ

Imperative Conclusive Adnominal

kake kaku kaku

> > >

Conclusive Adnominal

kai-tari kai-taru

> >

Conclusive Adnominal

kaka-u kaka-u

Conclusive Adnominal

kake

Imperative

kaku

Non past

kaitaru >

kaita

Past

> >

kakau =>

kakX>

Volitional

kaka-ilzu kake-ilzuru

> >

kakauzuru =>

kakXYZUru

Intentional

Conclusive Adnominal

kai-tara-u kai-tara-u

> >

kaitarau >

kaitarX>

Past conjectural

Imperative Conclusive Adnominal

akeyo aku akuru

> > >

akei

Imperative

akuru

Non past

Conclusive Adnominal

ake-tari ake-taru

> >

aketaru>

aketa

Past

Conclusive Adnominal

ake-u ake-u

> >

akeu=>

akyoo

Volitional

Conclusive Adnominal

ake-ilzu ake-ilzuru

> >

akeuzuru =>

akyoozuru

Intentional

Conclusive Adnominal

ake-tara-u ake-tara-u

> >

aketarau >

aketarX>

Past conjectural

and inflection. It was greatly simplified through early NJ, to comprise only a non-tensed provisional and a conditional, see 15 .1. In LMJ the past conditional was found both in its full shape, -(I)taraba, and in the abbreviated form -(I) tara.

(1) Provisional Conditional Concessive

Nonpast kakeba kakaba kakedomo

Past kaitareba kaitara(ba) kaitaredomo

Intentional kabazureba

Provisional Conditional Concessive

akureba akeba akuredomo

aketareba aketara(ba) aketaredomo

akyoozureba

kabazuredomo

akyoozuredomo

334

12 LMJ grammar

12.1.4

Conjectural

As mentioned above, the OJ/EMJ conjectural auxiliary -(a)m- changed to a flective -(a)u which only had volitional function with most verbs, hence the emergence of volitional forms of verbs (12.1.3.2). However, with existential verbs, and constructions involving existential verbs, OJ/EMJ -(a)m- only had conjectural function, and that also held for the new flective -(a)u when used with existential ar- and with forms built on ar-. Thus, EMJ ara-fi 'existconjectural'> LMJ arau > araa which was used in conjectural function, both as an independent existential verb, and in its grammaticalized uses to form analytic adjectival and regular copula forms and after verb gerunds to form an analytic stative. Similar developments hold for the EMJ stative conjectural -tara-fi > LMJ past conjectural -(I)tarau > -(I)taraa, and for the secondary adjectival and regular copula conjugations: -kara-fi > -karau > -karaa and nara-fi > narau > naraa. With ar- and the secondary adjectival and regular copula conjugations, the intentional, -(a)fize-, was also used, and was reflected in forms such as EMJ ara-fizu(ru) > LMJ arauzu(ru) > araazu(ru), and adjectival -karaazu(ru) and copula naraazu(ru), but with no discernible functional difference from the conjectural. In addition to the conjectural auxiliary -(a)m-, the extension ram- already in OJ provided an alternative, unambiguous expression of the present conjectural (3.6). In the course of the onbin sound changes the conclusive and adnominal ramu changed to ran ~ rail > early LMJ rau > late LMJ raa, which eventually became a particle due to the loss of the exclamatory as a distinct morphological category (12.1.2). It was used with the nonpast of verbs in general and with intentional -(a)uze- (-uzu raa), and with recent past -(I)te-; here the combination was univerbated as a past conjectural -(I)turaa, which continued in use after the loss of recent past -(I)te-. In late LMJ raa was frequent in the shomono texts and was also used in Esopo and Feiqe, and it became the main expression of the category of conjectural. In late LMJ there were thus two sets of expressions of conjectural which seem to have been more or less synonymous: (a) use of raa as a final particle (and in an inflectional past conjectural ending), and (b) a number of specific inflected forms, reflecting OJ ara-mu, see (2). In NJ, raa is no longer used, but has been functionally replaced by daro or its polite variant desha (both originating as conjectural forms of the copula) following a finite verb (kaku daro, kaita daro) or adjective form (akai daro), or directly following a nominal predicate (hon daro). In addition to daro (and other copula forms, such as de aro), adjectival copula -karo and past conjectural -(I)taro are still used, in old fashioned writing and productively in some dialects.

12.1 Verb morphology (2)

335 b.

a.

roo (< rau < rail < ramu) Verb, nonpast arAdjectival copula Copula Verb, past

(kaku) r::xJ aru r::xJ -ku aru raa (~ -karu raa) de aru raa -(I)turaa

a~

{< ara-u < ara-il < ara-mu)

araa -karaa naraa -(I)taraa

cNJ daro does not seem to be a direct reflex of a form like *dara-u, but rather an analogical formation. Part of the background for that is probably the extensive use of raa and also that the synchronic segmentation of the past conjectural, -(I)taraa, in late LMJ was not obvious: was it past -(I)ta + raa, conforming to the pattern with nonpast verbs (LMJ kaku raa); or was it an unanalysable single form (conforming to diachronic formations of past conjectural forms through the history of the language: OJ -{i)kye-m- > EMJ -(i)kem-, replaced by LMJ -(I)tu(-)rau > -(I)turaa, further replaced by -(I)tara-u > -(I)tarau > -(I)taraa > NJ -(I)taroo > cNJ -(I)taro). Note also in this connection early LMJ yarau (> LMJ yaraa) which originates in a contraction of EMJ ya ara-il, the kakari particle ya followed by ar- + conjectural. Yaraa is a particle with the meaning 'is it, I wonder?' used after nominal predicates and after verbs, both nonpast and past, and thus functionally corresponding to a conjectural particle and copula form. Yaraa could also have served as a model for analogical formation of a form such as daroo; further, yaraa is possibly the source of the modern Kansai copula ya which could have arisen in a reanalysis of yaraa as the conjectural form of ya, with further forms such as past yatta and negative ya na- being analogical formations.

12.1.4.1 Negative conjectural Negative conjectural was expressed by reflexes of the EMJ extension mazi(8.5): mazii which directly reflects the EMJ adnominal mazi-i and a further reduced form mai. Both forms were used in attributive and predicative function, though usually the latter. Like raa, mazii ~ mai changed from extension to particle in the course of the concl/adn merger. Mazii ~ mai followed the nonpast of consonant base verbs (e.g.fairu mai,fair- 'enter'); with vowel base verbs (including the passive and causative auxiliaries which belonged to the LB conjugation) it usually followed the base (uke mai, uke- 'receive'Jatasare mai, fatas-are 'fulfil-PASS'), or with se- 'do' and ko- the old conclusive (su mai, ku mai). Mai survives into cNJ, but mainly in conservative writing. Negative conjectural is instead generally expressed by dar8/desh8 following a negative form.

336

12 LMJ grammar

Table 12.5 Main forms of the negative at the end of the LM.J period leak-

ake-

Nonpast Past Past conjectural

nu nanda nandaroo

kakanu kakananda kakanandaroo

akenu akenanda akenandaroo

Infinitive Past infinitive Gerund Concessive Provisional Past provisional Conditional

zu nande ide nedomo neba nandareba zuwa, zunba

kakazu kakanande kakaide kakanedomo kakaneba kakanandareba kakazuwa, kakazunba

akezu akenande akeide akenedomo akeneba akenandareba akezuwa, akezunba

12.1. 5

Other auxiliaries

12.1.5.1 Negative Of the OJ auxiliaries discussed in 3.1.4.5, only the negative has remained an auxiliary through the history of the language, attaching to the a- stem of consonant base verbs and the base of vowel base verbs. Main forms at the end of the period are shown in Table 12.5. The negative inflects for many of the categories that verbs inflect for. Note that the paradigm of the negative was completely replaced in the standard language in NJ, see 16.3. The late LMJ negative paradigm is more suppletive than the EMJ negative paradigm (8.2.2). The nonpast directly reflects the EMJ adnominal and several other forms are direct continuations of the EMJ forms. However, the paradigm also includes a number of new forms. Interestingly, their etymologies are all unknown and they all went out of use during NJ in the common language (but are retained dialectally): The gerund -ide is similar to the EMJ negative gerund -de, but any derivational relationship is unclear. A number of negative past tense forms (past -nanda, past conjectural -nandar:J:J, past infinitive -nande, past provisional -nandareba) reflect a past negative auxiliary, -(a)nandar-, which came into use in the first half of LMJ. It has the basic paradigm shown in (3a) below. Alongside other auxiliaries it changed from an auxiliary to give a number of inflected forms in the second half ofLMJ, but its basic paradigm already shows signs of that process being under way. Past -nanda continued in use into the nineteenth century. Etymologically, -(a)nandar- seems to involve a combination of negative -(a)n- and the stative (past) auxiliary -(!)tar-, but the details of the composition of the form are entirely unclear. The secondary negative auxiliary -(a)zar-, whose basic paradigm is repeated in (3b), was frequent in EMJ and in the first half of LMJ; for example, the depiction of the three monkeys which 'see no evil, hear no evil, and speak no evil' (mizaro, kikazaro, iwazaro) involves a pun on the word saro 'monkey'

12.1 Verb morphology

337

and the negative -zaru. The adnominal -zaru came to be used in both adnominal and conclusive function as the two forms merged (12.1.2). Combinations with the reflex of the EMJ simple past adnominal -zassi (< zari-si) and with the new past tense zatta (< -zari-ta(ru)) were used through LMJ, but from the middle of the period started being replaced by -(a)nandar- which came into the language then. Use of -(a)zar- decreased during late LMJ and it is not reflected in the paradigm in Table 12.5. It was, however, retained in use in classical-flavoured writing. (3)

a. LMJ negative past -(a)nandar-

b. Secondary negative -(a)zar-

a- stem onbin stem

nandara-

zarazaQ-

Infinitive Conclusive Adnominal Exclamatory Imperative

nande (~ nandari) nanda nanda (~ nandaru) nan dare

zari zari zaru zare zare

12.1.5.2 Passive and causative Passive -rare- and causative -sase- remained in use, attaching to host verbs in the same way as in EMJ and belonging to the LB conjugation, that is, with the same forms as lexical LB verbs, e.g. nonpast -sasuru, -raruru; past -saseta, -rareta; volitional -sasyoo, -raryoo, etc.; especially in late LMJ, a number of irregular 'analogical' shapes of these two auxiliaries are found, e.g. -sasase- or -rarare-. Causative and passive basically had the same functions as in EMJ (8.4.1), with the notable exception that their exalting function became independent so that both passive and causative were now used to express respect on their own (cf. 12.7.1.2), as opposed to EMJ where they supported other respect morphemes. In late LMJ causative + passive, -sase-rare-, was often used to express a higher level of respect. Probably related to that are other multiple occurrences of these two auxiliaries seen in the texts, though an example such as ut-are-rare-sase-rareta 'was stricken.RESP' (with a double passive, -are-rare, expressing passive voice, and causative-passive, -saserare-, expressing respect) is very rare. 12.1.5.3 New auxiliaries A number of new auxiliaries came into use during LMJ. New auxiliaries in use at the end of the period are summarized in (4):

338

12 LMJ grammar

(4)

Potential Desiderative Evidential Evidential Polite

-e-(i)ta-(i)s:x>-(i)ge-(i)marase-

The ancestor of the cNJ potential auxiliary -re-makes its appearance in the second half of LMJ. It is found with consonant base verbs in the shape -e-, e.g. yom-e-nu 'cannot read' in the Shiki-sho from 1477, and initially had the same functions as the passive: passive, potential, and respect. -E- is usually said to have developed as a reduced variant of the passive -rare- and for example in the Shiki-sho we find both yom-e-nu and yoma-re-nu 'read-PASSNEG; cannot read'. Alternatively, -e- may have developed from the auxiliary verb -e- 'be able to'(< e- 'get'). During NJ this auxiliary became specialized as an independent potential auxiliary, but only recently in cNJ was it analogically extended to be used with vowel base verbs, e.g. cNJ tabe-re- 'be able to eat'. Several other new auxiliaries involve in their history a shift of a lexical item to grammatical function. Desiderative -(i)ta- (thought to derive from the adjective ita- 'intense, much, wonderful') became firmly established as an auxiliary during early LMJ, eventually replacing EMJ -(a)maosi-. As in cNJ, -(i)ta- had adjectival inflection and attached to the infinitive of verbs, e.g. early LMJ conclusive kaki-ta-si, ake-ta-si, adnominal kaki-ta-i, ake-ta-i. During LMJ, VERB-ta- expressed both 'want to VERB' and 'want somebody else to VERB', but in NJ specialized to mean '(I) want to VERB'. Already in LMJ -(i)ta- was used in combination with the evidential derivative -gar-: kaki-ta-gar- 'he seems to want to write, he wants to write'. From late LMJ, the early LMJ noun sau (which is thought to be a reduced form of either sama 'appearance, shape, state' or SJ sau <sail 'appearance, shape' (tEl, EMC *siaiJ!siaiJh 'look at, inspect /observe, judge the appearance') became used as an evidential auxiliary about appearance of state or imminence: 'looks/seems as if/as if about to', attaching to the infinitive of verbs and the stem of adjectives, forming an adjectival noun, i.e., combining with the copula: -(i)sau-nar- > -(i)sa:rna (>NJ -(i)soo-da), e.g. taka-sa:rna 'it looks tall', kaki-sa:rna 'he looks, seems about to go'. This is still used widely in NJ. In NJ soo-na further developed uses as a hearsay extension (15.3). Also two polite style auxiliaries have lexical sources (see 12.7.2): -(i)marase-, which is the ancestor of cNJ -(i)mas-, developed during late LMJ; it is from mawirase- 'give.HUM', a lexicalized causative of mawir- 'come. HUM, go.HUM'. Before then, -(i)saaraa was used as a polite auxiliary but had gone out of use by the end of LMJ; it is from sauraw- (a suppletive polite

12.2 Adjectives and copula

339

equivalent of the existential verb ar-) <early EMJ saburap-, said to be from OJ samorap- 'serve, be in attendance') The EMJ derivational morpheme -ge-(nar-) which derived an adjectival noun from adjectives (8.6), came to be used more widely as an evidential auxiliary, -(i)ge-(na) 'appears, seems', attaching to the infinitive of verbs and to the stem of adjectives. It also developed uses as an extension, with the same meaning, but following finite verb forms, see for example (7b) in 12.2.2.

12.1.6 Extensions Of the EMJ extensions (see 8.5), evidential nar- (sound, hearsay) went out of use after EMJ, and evidential mer- (sight) disappeared in the first half of LMJ. Present conjectural ram- and negative necessitive mazi- changed to particles as mentioned above (12.1.4). Necessitive be- was used through the period: there are tendencies already during LMJ to use nonpast bei/bee (< adnominal beki) and infinitive byoo (< beu < beku) as particles, but on the whole be- still at the end of the period had productive adjectival inflection. In addition to EMJ bekasi- (8.2.2, 8.5), a number of other analogical forms are found through the period, e.g. bera-nar-, besii-. The SJ noun EMJ yaii 'appearance, shape'(~ EMC *jiatf), which already in EMJ had conventionalized uses, early in LMJ acquired fully grammaticalized functions as an extension: early LMJ yau-nar- > late LMJ y:x~na (> NJ yoo-da), belonging to the class of adjectival nouns. Initially its uses were similar to goto- '(is) like', but eventually it became an evidential 'seems (as if), appears'. It is still used as an evidential in cNJ, yo-da, and its infinitive, yo-ni, has further been grammaticalized as a complementizer and a conjunctional particle.

12.2

Adjectives and copula

In EMJ the adjectival copula and the regular copula both had a highly suppletive paradigm of primary forms, as well as secondary, fully inflected verbal forms: adjectival copula -kar- and copula nar-ltar- which arose from lexicalization of phonological fusion of -ku ar- and ni/to ar- (8.2.1). By the end of LMJ the two paradigms of primary and secondary forms had, for both the adjectival and the regular copula, merged, or were in the final stages of merging, into one suppletive paradigm which reflects both primary and secondary forms and which incorporates some of the new categories for which verbs came to inflect, see 12.2.1 and 12.2.2 below. Like the verbs, the adjectival and regular copula at this stage both still retained direct combination with the negative auxiliary. As in EMJ, the frequent use of analytic forms, -ul-ku ar- and de/nite/ni ar-, continued, supplementing the new paradigms, using also

340

12 LMJ grammar Table 12.6 Late LMJ adjectival copula forms Finite Nonpast (Past) Conjectural Imperative

i (- karu) - kere

(katta) karoo- karoozuru kare

Non-finite Infinitive Gerund Conditional Provisional Concessive

u -ku ute- kute kuwa - kunba - utewa - kutewa kereba keredomo - utemo - kutemo

Negative

kara-n-

other existential verbs than the neutral ar-, particularly polite gozar-. Note that towards the end of the period the negative adjective na- had become regularly used as a suppletive negated form of ar- and that analytic negation generally used na-, rather than a negative form of ar- (such as aran), thus negative adjectival copula -ul-ku na-, negative copula de(wa) na-; cf. also 16.3.

12.2.1

Aqjectives

For the adjectival copula, the concl/adn merger left nonpast -i as the only reflex of the OJ conclusive and adnominal formants, (5a). With no reflex of the conclusive form, which is what distinguished the so-called ku and shiku adjectives (3.2.4), the difference between these two classes of adjective disappeared and was reflected only as a difference in stem shape, i.e., whether the stem ends in -si or not, (5b): OJ

(5) a.

b.

EMJ si ~ 0

LMJ

Conclusive Adnominal

si ~0 ki

Conclusive Adnominal

taka-si taka-ki

taka-si taka-i

> >

taka-i

Nonpast

Conclusive Adnominal

utukusiutukusi-ki

utukusiutukusi-i

> >

utukusi-i

Nonpast

> >

Nonpast

The main late LMJ forms and variants of the adjectival copula are given in Table 12.6. At this stage, the adjectival copula had most of the categories for which verbs inflect. However, past tense was usually formed analytically, e.g. -ul-ku

12.2 Adjectives and copula

341

atta or -ul-ku gozatta. Past -katta was not yet well established~ it was rare and is found mostly with na- 'not exist' which in LMJ seems to have been close to having been lexicalized as a verb nakar-, from the secondary conjugation (na-kar-). It never made the transition fully, but did exhibit forms which were not generally found for other adjectives such as nakatta or nakaru. The nonpast variant -karu reflects the EMJ secondary adnominal and was rare in late LMJ~ it was used in both attributive and predicative function, and also with the conjectural particles raa and mazii (12.1.4). Also imperative -kare reflects the secondary conjugation. -kere reflects the EMJ exclamatory and was used as a variant of the LMJ nonpast in agreement with koso. Morphologically -karaa ~ -karaazuru reflect combinations of the secondary conjugation -kar- with the EMJ auxiliaries conjectural -(a)m- and intentional -(a)iize- and thus correspond to the volitional and intentional in the verbal conjugations, but functionally -karaa ~ -karaazuru are both conjectural. Of the two infinitive variants, -u and -ku, the -u variant was used more, but both are found in the same environments, for example in analytic constructions with an existential verb or the negative adjective na-, e.g. past -u atta ~ -ku atta or negative -u na- -ku na-, or as adverbial or non-finite forms. The main exception seems to have been when used as a stem in the formation of the two conditionals where only -ku was used: -kuwa ~ -kunba~ of these, -kunba is more typical of kanbunkundoku. Like the verbs, the adjectival copula had new conditional and concessive forms built on the gerund: -utewa ~ -kutewa and -utemo ~ -kutemo. The adjectival copula still combined with the negative as -karan-, but increasingly negation was expressed by analytic forms, e.g. -ul-ku na-, as is the case in NJ. Note finally that the concessive adjectival form -keredomo already in late LMJ came to be used as an independent conjunctional particle keredomo 'but, however', with later frequent NJ reduced forms kedo, keredo.

12.2.2

Copula

Main inflected forms and variants of the regular copula at the end of LMJ are given in Table 12.7. The inflected copula tar-, which was restricted in use all along (8.2.3), was lost altogether outside fossilized expressions (e.g. in some adnouns such as kaku-taru 'certain'), but infinitive to remained. The forms in Table 12.7 reflect EMJ primary forms, EMJ secondary forms (from nar-) and new secondary forms which arose in late LMJ: dyar- (<dear-). The two secondary conjugations thus differ in whether they are built on the infinitive ni or on the gerund de(< nite). The nonpast forms in the LMJ paradigm are abbreviated reflexes of adnominal secondary forms, with loss of final -ru (see 12.4.1): na < naru and dya <

12 LMJ grammar

342

Table 12.7 Late LMJ copula forms Nonpast Attributive Conjectural Infinitive Gerund Conditonal Provisional Concessive Negative

na - dya (- dea- da) - nare no narY.J ni - to de - nite naraba (- dewa) nareba naredomo - demo nara-n-

dyant < de ant (and dea <de ant). Nonpast dais also thought to derive from de ant; it is a marginal form attested but a few times in shOmono texts and only included here because it is the form found in cNJ (15.2, 16.2). Although there was a tendency for dya to be used in main clauses and na adnominally, both could be used in either function, see (6) and (7). As in OJ/EMJ (and today in cNJ), LMJ had a distinct attributive copula form no, which was restricted in use and differed from the verbal OJ/EMJ adnominal and LMJ nonpast forms in being used only in adnominal function (cf. 3.3.1). A reflex of the EMJ exclamatory nare was used as a variant of the nonpast in agreement with koso until the end of the period (12.6.1.2.3). (6) a.

b.

(7) a.

b.

wast wa syotyoo no I TOP birds GEN 'I am king of the birds' (Esopo)

::J::l dya king COP.NONPST

kono koto wa asa-kara-nu TOP shallow-ACOP-NEG.NONPST this thing fusin dya fodo m suspicion COP.NONPST extent COP.INF 'because this is highly suspicious' (Esopo) kata-me na one-eye COP.NONPST 'the one-eyed deer'

sika deer

kono mono wa tsyu ga nao this person furthermore concern NOM TOP gozaru ge-na exist. POL EVID-COP.NONPST 'It appears that this person furthermore has some concern' (Feiqe)

12.2 Adjectives and copula

343

Naraba, nareba, naredomo reflect the EMJ inflected copula nar-, as do combinations with the negative auxiliary (12.1.5.1), such as nonpast naranu 'it isn't' or past provisional naranandareba 'as it wasn't'. Of the two gerunds de was used far more widely, but its source nite is still used in some texts, e.g. Feiqe. Etymologically de and nite are built on infinitive ni, but already from late EMJ, ni was getting more restricted and de or nite were used more generally as non-finite copula, 'is and ... '~ this tendency continued into LMJ where de also took over the function of being the form on which analytic forms were built: de ar- (which as mentioned gave rise to a short-lived secondary form: dyar-). The LMJ copula paradigm underwent significant changes in NJ, but the analytic forms were much as they are in cNJ, which for example uses polite negative dewa arimasen (or contractedja arimasen), or formal written style copula de ar-. In addition to the neutral existential verb ar-, other existential verbs were also used to form analytic copula forms, e.g. de gozar- which is also used in cNJ, especially as a superpolite copula de gozaimas-. While negation could be expressed by the negative auxiliary, naranu 'isn't', it was more usually analytically formed, using either a negated form of an existential verb, or the adjective na- which in LMJ became the regular suppletive negative form of ar- also in these constructions, e.g. dena- or dewa na-, as in cNJ (cf. 16.3). Severallexicalized forms reflect the secondary conjugation: nari-tomo 'be it, even' (< conclusive nari + conjunctional concessive particle tomo), naradewa 'only, limited to'(< negative gerund narade + wa), and the conditional particle naraba. 12.2.3

Summary of the development of regular and adjectival copula paradigms

The development of the morphology of the regular and adjectival copula through time is strikingly similar~ to some extent this is shared by the negative auxiliary. First of all, both regular and adjectival copula have through the attested history of Japanese had analytic constructions consisting of infinitive or gerund and an existential verb. This also holds for the negative in OJ, EMJ and early LMJ. In OJ, regular (3.3) and adjectival (3.2) copula and the negative auxiliary (3.1.4. 8) each had a paradigm of primary inflected forms, which was irregular, highly suppletive and, for the copula, very defective. These primary paradigms were supplemented by analytic forms, built on the infinitives -ku, ni/to and -zu. In EMJ both primary and analytic forms continued, with a few changes in the primary paradigms, but in addition lexicalized, fully verbal secondary paradigms emerged, which originated in the analytic forms (8.2). Finally, in late LMJ, the primary and secondary adjectival and regular copula paradigms merged into single suppletive primary paradigms whose forms reflect both the primary OJ/EMJ paradigms and the secondary EMJ paradigms

344

12 LMJ grammar Table 12.8 Development of the copula and adjectival copula paradigms EMJ/early LMJ

OJ

primary paradigms primary paradigms ni, no, to, tu; ~ ni, nile, de, no, to si, lei, ku, . . . si, lei - i, ku - u, . . .

/

lateLMJ/NJ primary paradigms da, na, no, ni, .. .; i, ku - u, katta, ...

~crya,

secondary paradig~

=~· tar-, cryar-

analytic forms analytic forms analytic forms nilto ar-; ------j ni/niteldelto ar- ------j de arku~

ku~

ku~

(12.1.5.1, 12.2.1-2). This final change coincided with and to a large extent paralleled the change of the verbal auxiliaries to inflected forms. The overall developments are summarized in Table 12.8.

12.3

Basic paradigms

The basic paradigms of the different verb classes are shown in Table 12.9. As opposed to the morphological system, the basic paradigms did not change much in the course of LMJ. The exclamatory was lost as an independent, distinct word form, but remained a stem used in the formation of the provisional and concessive verb forms and we now list it as a stem (' e- stem'), alongside the a-stem and the onbin stem of the consonant base verbs, but also for the vowel base verbs. This mode of presentation is an adaptation of the katsuyokei-system used in Japanese school grammar, with the forms slightly rearranged and the onbin stem added (cf. 3.4.6.3). For OJ and EMJ this presentation at once captures core inflected verb forms and stems, providing a basis for forming other forms (cf. 3.4.6), but for LMJ (and NJ) it is much less useful. It mainly has the latter function of providing a basis for the formation of morphophonologically more complex forms, but is in that respect more cumbersome than necessary, and for example the presentation of verbs in Vocabulario (10.2.2.3) is simpler and does the same job. Finally, if used diachronically, in particular comparing EMJ and LMJ, the presentation in terms of these basic paradigms gives the wrong impression of morphological continuity across these periods and does not capture the sweeping changes which the morphological system underwent between EMJ and LMJ. However, as mentioned before, this mode of presentation is ingrained in traditional Japanese scholarship so we include it for LMJ as well.

12.3 Basic paradigms

345

Table 12.9 LMJ basic verb paradigms N-irr

QD

QD

QD

Basic stem

sin'die'

mot'hold'

tor'take'

ar'exist'

a- stem onbin stem e- stem

sinasinsinure-

motamoQmote-

loratoQtore-

araaQ-

Infinitive Nonpast Imperative

sini sinuru sine

moti motu mote

tori toru tore

QD

QD

QD

Basic stem

kak'write'

lwg'row'

sas'pierce'

a- stem onbin stem e- stem

kakakaikake-

lwgalwilwge-

sasasaisase-

Infinitive Nonpast Imperative

kaki kaku kake

lwgi lwgu lwge

sasi sasu sase

QD

QD

QD

Basic stem

yom'read'

yob'call'

tow'request'

a- stem onbin stem e- stem

yomayou-- )'l?nyome-

yobayou-- yonyobe-

Iowalou-- toQtoe-

Infinitive Nonpast Imperative

yomi yomu yome

yobi yobu yobe

toi lou toe

LM

LB

LB

LB

Basic stem

ke'kick'

ake'open, (trans.)'

tae- (//taye//) 'cease, (v.intr.)'

kae- (//kaye//) 'exchange'

e- stem

kere-

akure-

tayure-

kayure-

Infinitive Nonpast Imperative

ke keru keyolkei

ake akuru akeyolakei

tae tayuru taeyoltaei

kae kayuru kaeyolkaei

are-

ari aru are

346

12 LMJ grammar Table 12.9 (cont.)

12.3.1

UM

UB

Basic stem

mi'see'

oki'arise'

e- stem

mire-

okure-

Infinitive Nonpast Imperative

mi miru miyo

oki okuru okiyo

s-irr

k-irr

Basic stem

s(e)'do'

ko'come'

e- stem

sure-

kure-

Infinitive Nonpast Imperative

si suru/su seyo/sei

lei kuru koyo/koi

Verb classes

As mentioned above, r-irr was no longer a separate class, but had become a part of the QD class. Other than that, the verb classes of LMJ are the same as those for EMJ. Innovating combinations of passive and causative with s-irr verbs arose, coexisting with the older combinations, for example of se- 'do': do.PASS sare(- non-innovating se-rare-) or do.CAUS sase- (- se-sase-). This, along with the persistence of the conclusive shape su (12.1.2), is usually interpreted to mean that the s-irr verbs were tentatively, at least by some speakers, reinterpreted as consonant base verbs, with the basic stem s-. In EMJ, there were two types of eve LB verbs, underlying //CVye// and underlying //CVe/1. The palatal glide in the first type only surfaced when followed by lui, never when followed by lei (because of the loss of /y/ before lei, see 7.3.2.2). This difference reflects three OJ sources of EMJ LB verbs in eve, as shown in (8), with developments including loss of /y/ (8a), 1-p-1 > 1-w-1 (8b) (7.3.1.1), and loss of /w/ (8b, c) (7.3.2.3). At the end of EMJ, type (a) was underlyingly //(C)Vye// (cf. conclusive tayu, infinitive tae), whereas types (band c) were //CVe// (conclusive kau, uu, infinitive kae, ue). However, by the end of LMJ all three types had become //(C)Vye//, see (9), as shown by uniform nonpast tayuru, kayuru, uyuru. This levelling of all EMJ eve verbs as //CVye// was probably prompted by the automatic phonetic palatal onglide of I.e/, manifested in the infinitive of all forms, regardless of diachronic source, as shown in (9).

12.3 Basic paradigms

(8) a.

347

Basic stem

Conclusive Infinitive

OJ C~eOJ taye- > EMJ tae- //taye-//: 'cease (intr.)'

tayu

OJ CVpeOJ kape- > EMJ kawe- > EMJ kae //kae//: kau 'exchange (v.tr.)'

b.

OJ CVweOJ uwe- 'starve' > EMJ ue //ue//:

C.

(9)

12.3.2

uu

LMJ CVe //CVye//

Nonpast

Infinitive

tae //taye// kae I /kaye/I ue 1/uye/1

tayuro kayuro uyuro

tae ([taie]) kae ([kaie]) ue ([uie])

tae ([taie])

kae ([kaie])

ue ([uie])

Onbin stems

Onbin stems were through LMJ regularly used with simple past adnominal -si, prohibitive na -(I)so, perfective -(I)te- and its reflexes (12.1.3), stative -(!)tar- and its reflexes, including past tense -(I)ta, and with gerund -(I)te. Both vocalic and consonantal onbin forms were found with -w, -b, -m base verbs, sometimes of the same verbs within one text, for example in the shomono; however, the Christian sources from the end of the period use almost exclusively the vocalic forms (see 16.1). Following the loss of phonemic nasality in moraic vowels and of the phonological rule of postnasal neutralization (11.1), morphophonological rules arose, like those of cNJ, which changed initial /t/ in the suffixes above to /d/ when used after onbin stems from -b-, -g-, -m-, -n-stems, i.e. those stems that gave nasal onbin in EMJ. Thus, the phonemic shapes of the gerunds did not change, but they were now produced by morphophonological rules which made reference to the final consonant of the basic stem of the verb, rather than to the final moraic consonant or vowel in the onbin stem. (10) a.

Basic stem

Onbin stem

yob- 'call' kog- 'row' yom- 'read' sin- 'die'

you-~

yon-

koiyou-~

sin-

yon-

Gerund

yoode koide yoode sinde

~ yonde ~ yonde

12 LMJ grammar

348

b. tow- 'ask' mot- 'hold' kak- 'write' kos- 'cross' tor- 'take'

12.3.3

tou- ~ toQmoQkaikoitoQ-

toote ~ totte motte kaite koite totte

Morphophonological rules

The monophthongizing sound changes described above in 11.5 gave rise to the morphophonological rules in (11), applying to verb and adjectival copula forms in -u, see (12)-(15). (11) a.

/-iu/ => 1-yuu/ l-eu/=> 1-yoo/ c. /-au/=> 1-:d d. l-ou/=> 1-oo/ e. 1-uu/ => 1-uu/

b.

(12)

Verbal u- onbin stem Onbin stem iu kanasiu

Gerund iu-te => kanasiu-de =>

yuute kanasyuude

b. ew- 'get drunk'

eu

eu-te =>

yoote

c.

kaw- 'buy' kam- 'bite'

kau kau

kau-te => kau-de =>

k:J:Jte k:J:Jde

d.

tow- 'ask' yom- 'read'

tou you

tou-te => you-de=>

toote yo ode

e.

kuw- 'eat' kum- 'draw (water)'

kuu kuu

kuu-te => kuu-de =>

kuute kuude

a.

Basic stem iw- 'say' kanasim- 'be sad'

Nonpast of -w- bases

(13) a. b. c. d. e.

Basic stem iw- 'say' ew- 'get drunk' aw- 'meet' tow- 'ask' kuw- 'eat'

Nonpast iu => eu => au=> tou => kuu =>

yuu yoo ::xJ [w:J:] too kuu

349

12.3 Basic paradigms

These rules still, with a few changes brought about by the NJ change of !':J':J! >fool (see 14.2), apply to u- onbin stems in dialects which use those stems, that is, particularly in Kansai; for example Kyoto kaw- 'buy' has the past koota (< b:Jta). However, they no longer apply to the nonpast of -w- bases, so for example aw- 'meet' today has the nonpast form au (not oo < :J:J). Note that cNJ yow- 'get drunk' is the result of a reanalysis of LMJ ew- on the basis of the monophthongized onbin stem and nonpast: eu-te 'drunk'=> yoote and eu 'gets drunk'=> yoo, reinterpreted as being from yow-. Likewise, iw- 'say' was reinterpreted as yuw- and listed in Vocabulario by the infinitive yui (although in standard NJ this verb today has the infinitive ii-, but other forms reflecting yuw-, e.g. nonpast yuu (spelled iu) or past yutta). The volitional fiective had the shape lui and, like its ancestor the conjectural auxiliary -(a)m-, attached to the a- stem of consonant base verbs and the basic stem of vowel base verbs. InNJthe shape of the volitional changed with vowel base verbs (see 15.1.1). Volitional -(a)u

(14)

Basic stem oki- 'arise' mi- 'see'

Volitional oki-u => mi-u =>

okyuu myuu

b.

ake- 'open' ne- 'sleep' se- 'do'

ake-u => ne-u => se-u =>

akyoo nyoo syoo

c.

kaw- 'buy' mot- 'hold' kak- 'write' kos- 'cross' yob- 'call' kog- 'row' yom- 'read' sin- 'die' tor- 'take' ar- 'exist' ko- 'come'

kawa-u=> mota-u => kaka-u => kosa-u => yoba-u => koga-u => yoma-u => sina-u => tora-u => ara-u => ko-u =>

ka:J:J ([kaw':J:]) mot:J:J kab:J kos:J:J yob:J:J kog:J:J yom:J:J sin:J:J tor:J:J ar:J:J koo

a.

d.

The adjectival copula infinitive -u is today widely used in Kansai dialects (sometimes with further changes in form), e.g. Kyoto taka- 'tall' + NEG => tako nai 'isn't tall'(< takoo nai < tab:Jnai). Standard cNJ generally does not use the adjectival -u infinitive, except in the so-called super-polite formation (a late borrowing from Kyoto, not used much today) which also incorporates

12 LMJ grammar

350

the monophthongization rules, e.g. utsukushi- 'beautiful' => utsukushil gozaimasu, taka- 'tall' => o-tako gozaimasu, and in a few forms lexicalized from that formation: ohayo (gozaimasu) 'good morning'(< haya- 'early') or omedeto (gozaimasu) 'congratulations'(< medeta- 'joyous'). Adjectival copula infinitive -u

(15) a. b. c. d. e.

12.4

Stem utukusi 'beautiful' be 'necessitive' taka 'tall' kuro 'black' usu 'thin'

Infinitive utukusi-u => be-u => taka-u => kuro-u => usu-u =>

utukusyuu byoo tab:J kuroo usuu

Existential verbs; ar-, i-, or-

The main, neutral verb of existence in Japanese is ar- which has been a prominent feature of the language since OJ, in more or less unchanged form and function. In addition to the uses as an existential verb, ar- has through the history of the language had important grammatical functions (3.4.2.1.1), namely (a) to form stative constructions and (b) to form extended analytic forms with the adjectival and regular copula and the negative auxiliary. Since OJ ar- has had a number of exalted or polite variants, e.g. OJ mas-, imas-, pab(y)er- (not phonographically attested in OJ, cf. 3.4.2.1 ), EMJ imase, opase(> owase-), LMJ odyar-, oryar-, goza(a)r- (cf. 12.7.2), which have also been used in the grammatical functions of ar-. NJ has two more existential verbs, i- and or-. In standard NJ, i- is used mostly with animate subjects, whereas ar- is used with inanimate subjects, and or- is a humble synonym of i- (although or- is used in some inflected forms instead of i-, e.g. infinitive ori). In other varieties of cNJ these verbs are used and distributed differently. For example, Kansai Japanese tends to use or- as the neutral verb of existence with animate subjects. Statives have through the history of Japanese been formed by a verbal infinitive or gerund+ ar- (or another verb of existence), and in OJ, EMJ and early LMJ also by stative auxiliaries (-yer- and -(!)tar-), both of which incorporate ar-. As mentioned above, the EMJ stative auxiliary -{!}tar-(< OJ -(i)te ar-) is reflected in the late LMJ and NJ past tense flective -(I)ta (12.1.3). In standard NJ statives are generally formed by gerund+ i- (kaite iru 'is writing, has written', usually abbreviated kaiteru), whereas gerund+ ar- is only used as a restricted resultative stative construction (kaite aru 'has been written'). Again, other varieties form statives differently, in Kansai Japanese mostly by gerund + or- (kaite oru, kaitoru). In some varieties of Japanese, statives

12.4 Existential verbs; ar-, i-, or-

351

include progressive function, but other dialects have separate expression of progressive, e.g. kakyoru ( <= kaki oru) 'is writing' in some varieties ofKansai, reflecting the OJ analytic progressive (3.1.4. 7.4). The two existential verbs NJ i- and or- derive from OJ wi- 'sit down, settle down' and war- 'be sitting', respectively. Kinsui's detailed study (2006) has shown (a) that these two verbs were not used as simple existentials until LMJ, and (b) that war- is the old lexicalized stative form of (an ancestor of) wi-, paralleling the regular morphological OJ stative in -yer-, e.g. sakyer- 'be blooming'(< *saki ar-), kyer- 'be wearing'(< *ki ar-), but reflecting a slightly different formation (e.g. *(w)u ar- or *wo ar-, attaching ar- to the diachronic root of wi-, not its infinitive). Thus, in OJ and EMJ wi- and war- were used as lexical verbs, wi- 'sit down' being the antonym of tat- 'rise, arise, stand up', and war- 'be sitting' the antonym of tater- 'be standing', the morphological stative of tat-. In addition, from EMJ a new regular stative was formed on wi- by the EMJ stative auxiliary to give wi-tar- 'be sitting' whose use gradually increased over war-. In OJ, war- was used to form progressives, e.g. tomosi war- 'be lighting' (3.1.4.7.4). This use continued in EMJ, where war- was also, though more rarely, used with the gerund rather than the infinitive. From EMJ progressives were also formed with wi-tar-, usually on the infinitive, but sometimes on the gerund (8.4.2.1). The fact that this use of war- and wi-tar- to form progressives precedes their use as free existential verbs, with animate subjects, suggests that the use in progressive constructions formed the basis for the reinterpretation of war- and wi- as existential verbs, that is to say, developments as shown in (16a).

(16)

a.

b.

Lexical verb

Stative lexical verb form

Use in progressive constructions

Existential verb

wi'sit'

war'be sitting down' wi-tar'be sitting down'

mi war'be seeing'

or'be'

mi witar'be seeing'

itar- > ita > i'be'

tat-er'be standing' pus-er'be lying down'

mi tater'be seeing' mi puser'be seeing'

tat'stand up' pus'lie down'

352

12 LMJ grammar

Note that the morphological statives of also tat- 'stand up' and pus- 'lie down', tater- 'be standing' and puser- 'be lying', in EMJ were used in progressive constructions, but that they were not reinterpreted as existential verbs, (16b). This offers strong support for the development proposed here, from a stative lexical verb to use in progressive constructions and then further to an existential verb (as opposed to a course of development where the use as existential verb precedes the use in progressive constructions). Such a use of stative positional verbs in progressive constructions is found in other languages, for example Danish, where ligge 'be lying', sidde 'sit, be sitting', sta 'be standing' are used as exemplified in (17), with the positional meaning bleached or lost. Outside such constructions these verbs are lexical stative verbs, which cannot be used as existentials. (17)

kmer rundt han ligger og around he is lying and drives 'he is driving around in his big car'

m

sm his

store big

bil car

Already in OJ, war- was univerbated, incorporating the original morphology as part of the lexical stem. As stative -(!)tar- changed from auxiliary to a ftective (12.1.3.2), EMJ wi-tar- > LMJ i-tar- gave late LMJ ita, and it appears to have been the form ita which was initially reinterpreted as a nonpast existential verb. In late LMJ there are uses of ita in nonpast existential function and this is still found in cNJ dialects in northern Japan, but ita was generally reformed to i-, presumably because the morphology remained transparent and ita does not conform to the basic stem shape of verbs. By the end of LMJ, i- was firmly established as an existential verb with animate subjects and was also widely used in stative constructions, mostly following the gerund, but sometimes following the infinitive. At that point, however, it had not entirely replaced ar- and its exalted and polite synomyms in these functions. Ar- was still used with animate subjects in late LMJ, alongside i-, but was eventually replaced by i- to give the distribution we find today. The specialized use of i- and or- as existential verbs with animate subjects is probably a reflection of their original semantics, which were mainly agentive and volitional, and of their use in progressives. The spread of the originally progressive construction with (w)ita(r-) to replace stative constructions in armay be thought to be related to the change to a past tense marker of the stative auxiliary -(!)tar- which incorporated ar- and also had analytic variants with ar- and its synonyms. Since war- started being replaced by the regular stative form wi-tar- in early EMJ, war> or- has been subject to various reinterpretations in different varieties, to a large extent in socio-linguistic terms, but or- has survived both in standard cNJ and in Kansai cNJ where, as mentioned above, it is in many varieties used as the neutral existential verb with animate subjects.

353

12.5 Pronouns and demonstratives

12.4.1

Loss of -ru

During LMJ a number of grammatical forms built on or diachronically incorporating ar- dropped the final -ro in reflexes of the adnominal form, see (18). Also the ending -ka stereotypical of Kyushu dialects, e.g. yo-ka 'is good', in the same way derives from the secondary conjugation of the adjectival copula: yo-karo > yo-ka. (18)

na COP.NONPST < naro COP.ADN dya COP.NONPST < dyaro COP.ADN <de aro da (very rare) COP.NONPST < *daru < dearo -(I)ta Past/Perfect tense < -taro STAT.ADN -(k)ke EMPH (cf. 12.1.3.1) < -kero MPST.ADN

12.5

Pronouns and demonstratives

Some terms of self-reference, address and 3rd person reference are shown in (19); these forms include demonstratives used for personal deictic reference. Among the 1st person terms, watakusi was earlier used for 3rd person (EMJ) and ore for 2nd person (OJ). (19)

1st person: watakusi, ware, ore, koti, kore, soregasi 2nd person: watono, nusi (originally 'owner, master'), onusi, onosi (< o-nusz), soti, konata, sonata, nandi, onore, kisyo (SJ 'noble, august place'), sonoj:J:J (f:J:J > foo > cNJ h8 'direction, side') 3rd person: kare, are, yatu, aitu, soitu, kore, sore Interrogatives: tare, dore

am

By the end of the period the system of demonstratives and interrogatives was close to the one we find in NJ. (20) Nominal Modifier Location Location/ direction Location/ direction Manner

Proximal kore kono koko koti/kotira

Mesial sore sono soko soti

Distal are ano asoko (/kasiko) atilatti

Interrogative dore dono doko doti

konata

sonata

anata

donata

kau > b:J

sau > S:J:J

d:J:J

12 LMJ grammar

354

The demonstrative fonns are parallel, but the most remarkable change is the levelling of the interrogative forms to confonn to the demonstratives; the interrogative adverb daa is from the very end of the period. The interrogative pronoun remained tare, however, and only changed to dare in late NJ; the use of donata as an interrogative pronoun is from early NJ. Anata was used for 3rd person reference in NJ, but changed towards the end of the eighteenth century to 2nd person reference.

12.6

Syntactic changes

Two sets of major syntactic changes have taken place in Japanese. One comprises the loss of the concl/adn distinction and includes also the loss of kakarimusubi (12.6.1) and the emergence of no as anominalizer (12.6.1.3). The other is the change of ga from a genitive to a nominative case particle (12.6.2). These changes are mainly manifested through the LMJ period, but are initiated or anticipated in late EMJ.

12.6.1

Loss of the distinction between conclusive and adnominal

As set out above, the loss of the distinction between conclusive and adnominal was an important part of the changes which took place in the morphological system. However, the distinction was functionally syntactic and the loss of it also fonns part of the complex of syntactic changes which took place during this period. The main functions of the conclusive and adnominal fonns in OJ and EMJ were as follows (cf. 3.1.3.1): (21) a. b. (22) a. b. C.

d. e.

Conclusive predicate in declarative main clauses with extensions, sentence final particles, and some conjunctional particles Adnominal predicate in adnominal clauses a nominalized fonn the predicate in exclamative or interrogative main clauses musubi predicate in kakari-musubi with zo, namu, ka, ya with some conjunctional and other particles

The concl/adn merger took the course of the adnominal being used increasingly to conclude declarative main clauses from late EMJ. This eventually resulted in the (descendants of the) adnominal taking over the functions earlier associated with the conclusive. This was not limited to the nonpast, but held also for the past, intentional, volitional and past conjectural, that is to say, all

12.6 Syntactic changes

355

the inflected fonns which arose as a result of the concl/adn merger (cf. 12.1.3.2). These fonns thus at the end of LMJ had the following main functions: (23) a. b. c. d.

predicate in all types of main clauses (except imperative) predicate in adnominal clauses a nominalized fonn use with extensions and particles

However, it is interesting to see that new distinct morphological marking of some of the functions in which the adnominal was used distinctively from the conclusive was acquired, or at least attempted, at the same time as, or after the concl/adn merger, namely when used as the predicate in kakari-musubi and exclamatives and interrogatives (12.6.1.2) and in nominalizations (12.6.1.3). 12. 6.1.1 Adnominal clauses As in NJ, by far the most common type of adnominal clause in premodern Japanese was as shown in (24), where a head noun is directly modified by a clause. This has been the main type of adnominal clause through the history of Japanese, but as opposed to NJ, premodern Japanese used the distinct adnominal verb fonn in adnominal clauses. The concl/adn merger thus meant the loss of the expression of the difference between a predicate in an adnominal clause and in declarative main clauses. However, there was from OJ onwards no segmental distinction among the QD verbs (or from EMJ onwards UM and LM verbs) between conclusive and adnominal forms, and as NJ shows, the absence of a conclusive versus adnominal distinction poses no parsing problems of adnominal clauses for speakers. In (24) there are two embedded adnominal clauses; the first has its predicate (su nant 'said to do') in an identifiable adnominal fonn, whereas the second does not, because the verb is a QD verb (ip- 'call') whose adnominal and conclusive were (segmentally) identical.

(24)

[wotoko mo su naru] [nikki to ipu] mono man ETOP do.CONCL EVID.ADN diary COMP call.ADN thing 'The thing called diary which men are said to keep' (Tosa)

12.6.1.1.1 Complementizers Japanese adnominal constructions generally do not have complementizers, i.e. linking elements, between an adnominal clause and a head noun. However, constructions such as (25) with a genitive particle between a modifying clause and a modified noun are found through premodern Japanese, although they are considered ungrammatical in NJ.

12 LMJ grammar

356

(25) a.

b.

c.

d.

OJ [ wagimokwo ni mise-mu] ga tame my.beloved DAT show-CONJ.ADN GA sake 'In order to show (them) to my beloved' (MYS 19.4222)

ni COP.INF

[taye-mu] no kokoro end-CONJ.ADN NO heart 'The intention to end it (our relationship)' (MYS 12.3071) EMJ [kabakari yononaka 0 no this COP.ADN world ACC no omowi-sute-mu] kokoro think-discard-CONJ.ADN heart NO 'The intention of abandoning this world' (Genji: Agemaki) Early LMJ [tokoro o omoi-sadame-zaru] ga yue ni place ACC think-decide-NEG.ADN GA reason COP.INF 'Because I hadn't settled on a (permanent) place' (Hojola)

In such constructions ga or no can be said to function as complementizers. 2 An important feature of this construction is that it was used only in 'gapless' adnominal clauses, as in the examples in (25), that is to say, noun modifying clauses where the modified noun has no syntactic function within the adnominal clause. In OJ there are not many phonographically attested examples, although the reading tradition of logographically written text posits quite a few more. In EMJ and early LMJ, the construction was used in small but steady num hers. It is not now possible to present a detailed account of the distribution and development of this construction, but the following tendencies may be noted: (a) in OJ, EMJ and early LMJ the construction- although never frequent- is clearly found in more examples than can be attributed to error; (b) in these periods, it is used more with ga and less with no as the particle between the clause and the noun; and (c) it is restricted to gapless constructions. ' There are other examples of adnominal clauses in OJ with a particle between the adnominal clause and its head noun such as (i), but it is doubtful whether ya (or similarly used i si, ya si, yo), which is usually glossed 'emphatic' in such examples, can be said to function as a complementizer here.

i.

[wotomye no nasu] ya ita-two maiden GEN sleep.RESP EMPH plank-door 'The door behind which the maiden is sleeping' (KK 2)

357

12.6 Syntactic changes

From the second half of LMJ two changes take place: the frequency of use of this construction increases, and no comes to be used regularly instead of ga, e.g. (26a). Although saying it was characteristic of written language, Rodrigues (Arte, p. 506) describes this use with nouns such as yoxi /yosi/ 'manner, way', aida 'while', yuye /yue/ 'reason', tocoro 'place', giving examples such as (26b). (26) a.

b.

Late LMJ [kuru o tanomu] no kari-no-tamadusa come ACC ask NO goose-GEN-letter (=letter) 'The letter asking me to come' (ShingoshUi wakashU, 1384) "xitagono tocoroni" [sitag~~] no follow.NONPST NO 'When following'

tokoro place

m COP.INF

Both of these changes in use are traditionally ascribed to a new kanbun-kundoku practice in the Zhu zi C*-1-) school of Confucianism, in which Chinese ~ was rendered as no, also in these constructions where the Japanese syntax does not require it~ an early precursor of this view was held by Moto'ori Norinaga, who also denounced such constructions as incorrect and uncharacteristic of the Japanese language. The idea is that use of the construction would have spread from the use in this school of kanbun-kundoku into common (written) language. However, as we saw above, the construction has been in use through all of premodern Japanese, and it is perhaps more likely that use of the construction increased through late LMJ in response to the merger of the distinction between the conclusive and adnominal verb forms. The construction was, however, never fully generalized and it is today considered ungrammatical. Even so, an informal internet search easily turns up cNJ examples such as (27a). It is also noteworthy that adnominal clauses with a genitive particle used as complementizer are regularly produced by children in the course of their acquisition of Japanese as a first language (as well as by second language learners), but here not limited to gapless adnominal clauses, e.g. (27b) by a 2+-year-old girl. (27) a.

[omoshiroi burogu ni suru] interesting blog COP do m wa COP.INF TOP 'in order to make an interesting blog'

no NO

tame purpose

12 LMJ grammar

358

b.

[chigau] no kami be. different NO paper 'a different piece of paper'

12.6.1.1.2 Tokoro-no A different type of complementizer is found in (28), with tokoro-no between adnominal clause and head noun. As opposed to adnominal clauses used with ga/no, tokoro-no can be used with relative clauses, that is, adnominal clauses with a syntactic gap, as shown in (28). The use of tokoro-no as a complementizer originates in kanbun-kundoku (cf. 9.1. 7) and seems only to have been used in written language, including its vocalized versions. It was revived in NJ in kanbun-kundoku-like renditions of Dutch texts (cf. 17.3.3) and may still be used in written language.

(28)

[0; tatekome-taru] tokoro-no to; close-STAT.ADN TOKORO-NO door 'The door which had been closed' (Take tori)

12.6.1.2 Loss ofkakari-musubi The kakari-musubi construction gradually declined in the course of LMJ and had all but disappeared from the language at the end of the period, when it was still found to a limited extent with koso, but no longer with namu, ya, zo, ka. Kakari-musubi is today only retained in a few mainland and some Ryukyuan dialects. Of the kakari particles, ka (interrogative) and zo (exclamative) are widely used as sentence final particles in cNJ, and koso is used as an emphatic particle~ koso and zo are also lexicalized in common expressions like yookoso 'welcome'(< yo-u< yo-ku 'good-INF' +koso) and doozo 'please~ here you are'(< doo 'how'+ zo). The loss of kakari-musubi is sometimes presented as a direct result of the concl/adn merger, but that is probably somewhat simplistic. First, that would leave unexplained the fact that kakari-musubi with koso was also lost (even if that happened a little later than the loss of kakari-musubi with an adnominal musubi predicate). In that case it seems clear that the loss of the distinct exclamatory verb form was a direct result of the loss of kakari-musubi with koso, for by the beginning of LMJ use in kakari-musubi with koso was the only function of the exclamatory. Second, there are mainland dialects (e.g. in Wakayama) which maintain a concl/adn distinction, but in which kakarimusubi is also lost. Third, any view of causal relations between the concl/adn merger and the loss of kakari-musubi depends on the view taken of what kakari-musubi actually was. If kakari-musubi were viewed as a simple agreement relation between one form (a kakari particle) and another (an adnominal or exclamatory musubi predicate), then naturally the loss of the distinctive predicate form would mean the end of that relation. However, we saw above

12.6 Syntactic changes

359

(8.9.3) that the fonn of the musubi predicate was not the defining feature of kakari-musubi: in many cases the musubi predicate did not have a distinct shape, namely, if it was in a subordinate clause, had a nominal predicate with zero copula, or the predicate was a QD or UM verb (which had no (segmental) distinction between conclusive and adnominal). If, on the other hand, as on Watanabe's analysis (8.9), kakari-musubi was basically a movement operation, then the loss of movement equalled the loss of kakari-musubi. This is what Watanabe suggests, saying further that this took place in OJ, leaving behind a simple agreement relation between kakari particle and predicate fonn which gradually disappeared because it was unmotivated and vacuous after the loss of movement. Quinn's functional account (8.9.2, 8.9.3) invests the kakari particle and the shape of the musubi predicate with separate, but combining functions, and on that view, the concl/adn merger should not affect the function of the kakari particle. It does seem intuitively clear that the loss of kakarimusubi and of the conclladn distinction fonn part of the same complex of changes, but the precise nature of the relation is difficult to establish. Below we discuss the possibility that the exclamative and interrogative functions of the adnominal fonn retained some morphological marking also after the concl/ adn merger (see 12.6.1.2.2). In any case, it is clear that the kakari-musubi construction -whatever the fonnal analysis of it- was an important focus construction in OJ and EMJ, and that the loss of it is among the most important syntactic changes to have affected Japanese. 12.6.1.2.1 Namu and ya Namu was lost entirely early in LMJ and is no longer found in the materials from the end of the period. Ya largely replaced ka inside yes/no questions through EMJ and was used through LMJ, but seems gradually to have been reinterpreted as expressing uncertainty, 'I wonder', rather than simply indicating yes/no questions~ through LMJ the function of indicating yes/no questions was taken over by sentence final ka. Ya was at the end of the period used mostly in combination with various modal fonns of ar-. Some of these combinations were lexicalized: Early LMJ yaraii (> LMJ yar:J:J) is a copula like particle, 'is it, I wonder?', originating in a contraction of ya ara-ii (< OJ ara-mu) 'exist-CONJ'. Yar:J:J is possibly the source of the modem Kansai copula ya (see 12.1.4). Early LMJ yaraii has also given the particle yara, still used in NJ with the meanings 'and so on~ I wonder' and also with wh- words to fonn existential quantifiers (doko-yara 'somewhere'). In addition, ya had by the end of LMJ developed uses as a particle 'and, or, or the like', often used together with nando or nado 'such as, etc.'. 12.6.1.2.2 Ka and zo The use of both zo and ka changed in the course of LMJ. Zo is only and ka mainly used as sentence final particles at the end of LMJ, e.g. (29)-(31). The main pattern shown in these examples, which are all

12 LMJ grammar

360

taken from Esopo, is that zo is used in exclamative main clauses (29), and in wh- questions (30), whereas ka is used in yes/no questions (31). This use of zo is for example also found in the shomono materials. (29) a. kurusyuude kuyamu wa tikusy~~ no waza suffer.GER regret.NONPST TOP beast GEN job 'suffering and regret is the burden of the beast!' (Esopo) b.

(30) a.

ookame no kuru wolf GEN come.NONPST 'The wolf is coming!' (Esopo) ware wa doko I TOP where 'where do I go?' (Esopo)

e ALL

zo ZO yuku go.NONPST

b. nandi ware o-ba tare who you I ACC-TOP omoo zo think.NONPST ZO 'Who do you think I am? (Esopo) c.

(31)

zo ZO

nandi wa naze m sita you TOP why COP.INF tongue b~te o-ba kuru buy.GER come.NONPST ACC-TOP 'Why did you only buy tongue?' (Esopo) kisyo wa yakusoku wa wasureta you TOP promise TOP forget.PST 'Did you forget your promise?' (Esopo)

zo

zo

to COMP

bakari only zo

zo ka KA

This distribution suggests that one important function of the adnominal verb form, namely to mark sentences as exclamative or interrogative, was taken over after the concl/adn merger by two former kakari particles, first of all zo, but in yes/no questions by ka. The use in (32) is quite rare, but noteworthy in this connection: it is a wh- question, marked by zo, but in addition, the whword is marked by ka, as opposed to the usual pattern shown in (30). (32) can thus be thought of as a late reflex of the OJ and EMJ kakari-musubi pattern in which wh- words were often marked by ka (more so in OJ than in EMJ) correlating with distinct sentence final morphology, in OJ and EMJ the adnominal form, but in late LMJ the particle zo. This suggests that zo was at once nominalizing and predicating~ zo could also predicate nominals as in (29a). On this interpretation, expression of the functional differentiation between

12.6 Syntactic changes

361

declaratives on the one hand and exclamatives and questions on the other was maintained also after the concl/adn merger. tarebito ka s1 ni who KA teacher COP.INF 'Who shall we make the teacher?' (Esopo)

(32)

syoo do.VOL

zo zo

The use of ka as a general sentence final interrogative particle in NJ is quite different from the LMJ pattern of distribution of ka and zo exemplified above and is thought to be an eastern dialect feature which spread after the ascendancy of the language of Edo/Tokyo. Note finally that the use of ka after interrogatives to form existential quantifiers (dare-ka 'someone', nani-ka 'something', etc.) is a NJ development which took place after ka entirely lost uses such as that exemplified in (32). 12.6.1.2.3 Koso As opposed to the other kakari particles, koso is still used today as an emphatic particle, although to a limited extent. In mainland dialects that retain kakari-musubi, this involves reflexes of koso. At the end of the LMJ period koso is still used in kakari-musubi focus constructions. It is not frequent, but it is found in both Esopo and Feiqe, e.g. (33a) which is an unambiguous focus construction and which has an exclamatory variant in the predicate (see 12.1.2 about the exclamatory). As mentioned above (12.1.2), the predicates used with koso in Esopo are mostly grammatical forms belonging to or reflecting r-irr verbs, but there are exceptions, even with those morphemes, where the predicate is not in the exclamatory variant. Thus, the use of koso in kakari-musubi at the end of the LMJ period represents the final stage in the loss of kakari-musubi. (33b) is an example from the first half of the sixteenth century (1534) where the predicate correlating with koso is not in the exclamatory.

(33) a.

sore o-ba Esopo koso nusunde that ACC-TOP Aesop steal.GER KOSO tabete gozare eat.GER exist(POL).EXCL 'It is Aesop who has stolen and eaten that' (Esopo)

b. kono kotowari wa koso kami god this reason TOP KOSO omoeba siru to know.NONPST think COMP 'If you think that it is god who knows the reason for this' (Shiganikkai)

12 LMJ grammar

362

12.6.1.3 Nominalizations; emergence ofnominalizing no In addition to the uses in adnominal clauses and in kakari-musubi, the OJ/EMJ adnominal functioned to form headless nominalizations (cf. 3.1.3.1), see (34), (35) and (36). Of these, examples like (34) are sometimes analysed as 'internally headed (or circum-nominal) relative clauses', with the head noun represented inside, but not after the modifying clause, so that in (34a) kiku no pana would be interpreted as a head noun modified by uturopem 'which has faded', or in (34b) tori would be interpreted as a head noun modified by two clauses, pasi to asi to akaki 'whose beak and feet are red' and sigi no opokisa nam 'which is the size of a snipe'. Examples like (35) have been analysed as 'headless relative clauses', meaning 'that which ... , the one who ... ', which have representation of a head noun neither after nor inside the adnominal clause. As mentioned in 3.1.3.1, such constructions were rare in OJ and almost exclusively used in pseudo-cleft constructions like (34a), but from EMJ the construction came to be used more widely, e.g. (35b). It has been proposed that such analyses of the constructions in (34) and (35) violate basic notions of relative clauses as being noun modifiers (Shibatani, p.c.). We here simply treat them neutrally as a subgroup of headless nominalizations. The pattern exemplified in (36) is often referred to as a complement clause. Use of the adnominal in that function was rare in OJ, which generally used the nominal form (cf. 3.1.3.3), but it became frequent in EMJ.

(34) a.

b.

[kiku no pana no chrysanthemum GEN flower GEN uturop-eru] wo worite fade-STAT.ADN ACC break.GER 'picking some faded chrysanthemum' (1se 18) [[ siro-ki tori no pas1 to white-ACOP.ADN bird beak and GEN to Slgl as1 no aka-ki] feet and red-ACOP.ADN smpe GEN opokisa naru] SIZe COP.ADN midu no upe ni asobitutu lWO water GEN top frolic.CONT DAT fish WO kupu ACC eat.CONCL 'A white bird which has a red beak and feet and is the size of a snipe was eating fish while frolicking on the water.' (1se 9)

363

12.6 Syntactic changes

(35) a.

b.

(36) a.

b.

[kadi no oto suru] pa GEN sound do.ADN TOP oar ama-wotomye kamo Q fisher-girl 'the ones making the oar-sounds, is that the fisher-girls?' (MYS 15.3641) [Usiromi to iu] call.ADN Ushiromi COMP naga-ku wokasige long-ACOP.INF beautiful 'As the one called Ushiromi had beautiful' (Ochikubo)

wa kami TOP hair nareba COP.PROV long hair and was

[imizi-u naku pito aru] terrible-A COP .INF person exist.ADN cry.ADN wo kikitukete ACC hear.GER 'hearing that there was a person who was crying terribly' (Ise 6) [yo pito-yo s1ra-nu koto know-NEG.ADN night one-night thing yon utipiki-tamawi-turu] m koso DAT depend torment-RESP-PERF.ADN FOC ito warina-kari-ture very unbearable-ACOP-PERF.EXCL 'It was unbearable that you tormented me through the night because of something I didn't know about' (Ochikubo)

In late LMJ, after the concl/adn merger, the inflected forms which resulted from the merger (the nonpast, past, volitional and intentional) came to be used in these nominalizing functions, as in (37):

(37) a.

sukasu [tiisai ko naku] o no cry calm.down small child GEN ACC fawa kamaete tote sono nakaba mother really! PURP that cry.COND ookame m yar:x:Jzu to yuu wolf DAT give.INT COMP say 'In order to calm down a small child who was crying, his mother said "if you keep crying, I'll give you to the wolf'.' (Esopo)

12 LMJ grammar

364

b.

[sisi no lion GEN no COP.ADN '(What) the (Esopo)

yuuta] wa ware wa arufodo say.PST TOP I TOP all ~ kedamono no nareba ... animal GEN king COP.PROV lion said: "As I am the king of all animals ... "'

c.

[kurusyuude kuyamu] wa tikusy:>:> suffer.GER regret.NONPST TOP beast no waza zo GEN job ZO 'It is the job of the beast to suffer and feel regret!' (Esopo)

However, in early NJ no emerged as a nominalizer which came to be used in these functions. (38) contains two early examples: (38) a.

[soregasi ga suite yomu] no I NOM like.GER read.NONPST NMLZ wa seisuiki o suite yomu TOP Seisuiki ACC like.GER read.NONPST 'What I like reading is (I like reading) the (Genpei) Seisuiki' (Kyogen-ki, c. 1660)

b.

[fara no tatu] no wa waru1 stomach GEN rise.NONPST NMLZ TOP bad.NONPST 'It is bad that he is angry' (Kabukijiihachibanshii, c. 1700)

Thus, nominalizing no took over the nominalizing functions of the OJ/EMJ adnominal and came to be used in all contexts where the OJ/EMJ adnominal formed headless nominalizations, see (39)-(41), but only after a hiatus with no distinct morphological expression of these functions, between the concl/ adn merger and the emergence of no as a nominalizer. In formal syntax, NJ nominalizing no is often given entirely different analyses in different contexts: as a pronoun in (40b), and as a (verb selected) complementizer in (41b). However, formal syntactic analysis does not agree about whether no in examples such as (38b), which are often analysed as internally headed relative clauses, should be interpreted as a pronoun or as a complementizer. (39) a. b.

[kiku no pana no uturop-eru] wo worite (=34a) Taroo wa [ringo ga sara no Taro TOP apple NOM plate GEN ue m atta] no o totta top DAT was] NMLZ ACC took 'Taro picked up an apple which was on a plate'

12.6 Syntactic changes

(40) a. b.

(41) a. b.

365

[Usiromi to iu] wa kami naga-ku wokasige nareba (=35b) [kino katta] no yesterday bought NMLZ 'I ate the one I bought yeserday'

o

tabeta ate

ACC

[imizi-u naku pito aru] wo kikitukete (=36a) [kino kita] no yesterday came NMLZ 'I forgot that I came yesterday'

wasureta forget.PST

0

ACC

There is a sizeable literature which presents different scenarios for the development of these nominalizing functions ofNJ no. The pronoun-like use of no, as in (40), predates the complementizer-like use (Kinsui 1995), but it has been pointed out by Wrona (forthcoming) that there is little need to posit hypothetical developmental steps between these functions if they are considered subfunctions of a general nominalizing function. As for the emergence of no as a nominalizer, Wrona points out that no emerges as a nominalizer after the increase in the use of no as a complementizer in adnominal clauses (see 12. 6.1.1.1 ), proposing that this use as complem entizer in adnominal clauses formed the basis for the generalization of no as a nominalizer.

12.6.1.4 Summary ofthe developments ofthefunctions ofthe Old Japanese/ Early Middle Japanese adnominal The expression in late LMJ and NJ of the functions which the OJ/EMJ/ early LMJ adnominal had can be summarized as in (42). Overall expression of the functions of the adnominal has been replaced by the final particles zo, ka (12.6.1.2) and by nominalizing no (12.6.1.3). It is conspicuous that the only function of the OJ/EMJ adnominal which does not have distinct morphological expression in late LMJ and NJ is the adnominal function, apart from the marginal use of no as a complementizer in ungapped adnominal clauses (12.6.1.1.1). (42)

LMJ and NJ expressions of the functions Late LMJ kakari-musubi (zo) exclamatives zo wh- questions zo yes/no questions ka adnominal clauses (other) nominalizations

(no)

of the EMJ adnominal NJ

zo ka ka (no) no

366

12 LMJ grammar

12. 6.2

The genitive particles, subject marking and the emergence of a nominative case particle

In OJ ga and no were genitive particles, used with some morphological, syntactic and semantic specialization (see 3.7.1.1.1). Both ga and no were in OJ and EMJ used to mark subjects of adnominal clauses, other subordinate clauses, and exclamative or interrogative main clauses (including in kakarimusubi, see 8.9). In OJ, EMJ and early LMJ ga was much more restricted than no as a genitive, both in adnominal and in subject marking function. In OJ ga was mainly used with pronouns and nouns referring to hum an beings, and through EMJ and early LMJ ga became increasingly restricted to use only with pronouns and nouns referring to 1st and 2nd person and with personal proper nouns. On the other hand, only ga, not no, was used to case mark headless nominalizations (as reflected in the development of ga into a conjunctional particle, cf. 8.7.2), and ga was used as a complementizer more than no (see 3.7.1.1.2, 12.6.1.1.1). A final minor context where ga was used rather than no was in fractions, e.g. sanbun ga iti 'one third'; this usage is found at least from late EMJ. No, by contrast, had no restrictions with regard to the type of noun it could mark, both in adnominal and in subject marking function, and through EMJ and early LMJ the use of no to mark subjects expanded. However, no did not make the full transition to marking subjects of declarative main clauses. However, at the end of LMJ the situation is markedly and remarkably different: ga had entirely lost its productive adnominal uses (being almost exclusively used to modify nouns in idioms and in lexicalized possessive pronouns such as waga 'mine') and also its function as a complementizer, but had expanded its use as a subject marker dramatically. Other than its independent use as a conjunctional particle, ga was by the end of LMJ only used productively to mark subjects, both in declarative main clauses, e.g. ( 43), as well as in other clauses. Thus, ga may be thought by the end ofLMJ to have completed a shift from a severely restricted genitive to nominative case particle, although ga even today in cNJ remains restricted in declarative main clauses, marking only focussed subjects and subjects in existential or presentational sentences. The emergence of a nominative case particle may well be considered one of the few major syntactic changes to have taken place in the attested history of Japanese. (43)

Amonia to yuu sato call village Amonia COMP 'There is a village called Amonia'

ga NOM

odyaru exist. POL

367

12.6 Syntactic changes

No, on the other hand, remained a genitive particle. In addition to its adnominal functions, no was still used to mark subjects in subordinate and nominalized clauses, e.g. (44). The circumscription of the use of no to mark subjects has continued into contemporary NJ where no still marks subjects in short adnominal clauses, but not in other subordinate clauses. (44) a.

b.

stsl no yuuta wa: ware wa arufodo lion GEN say.PST TOP I TOP all no kedamono no ~ nareba ... COP.ADN animal GEN king COP.PROV 'The lion said: "As I am the king of all animals ... "' (Esopo) ookame no kuru wolf GEN come.NONPST 'The wolf is coming' (Esopo)

zo zo

Conversely, no in late LMJ became increasingly used as a complementizer in certain adnominal clauses (see 12.6.1.1.1) and also developed its general nominalizing function (12.6.1.3). The development of the functions of OJ genitive no and ga are summarized in Table 12.10. The shifts in use of ga and no between early and late LMJ are remarkable, but their course has not been charted in detail. It appears as if no and ga around the middle of LMJ swapped or flipped important syntactic functions between them: subject marking and use as complementizer. It is a major challenge facing future studies in Japanese diachronic syntax to provide a plausible account of this, as well as of the development of the functions of ga and no in different dialects (where they have the same or similar functions, but distributed differently over the two particles, cf. also 3. 7.1.1.1).

12. 6.2.1 Socio-linguistic differentiation of no and ga Through EMJ and LMJ no and ga are famously said to have been differentiated in socio-linguistic terms, so that ga was deprecating or humble, whereas no was deferential, respectful or neutral. Rodrigues is quite explicit about this in the description in Arle of ga and no, which are both included among the particles expressing the nominative ('que servem ao Nominativo'Y and are given as the only two genitive particles. ga is used in careful speech and it indicates that what is referred to by the noun preceding it is deprecated. It is used with a first-person pronoun or a third-person pronoun referring to a low-ranked person. Also it is used to slight or deprecate other persons.

3

The other articles listed as serving that function (that is, marking a phrase that can be thought of as the subject) are wa, yori, kara, ni, niwa, ni oitewa.

368

12 LMJ grammar

Table 12.10 Summary of the development of the uses ofga and no

OJ

early EMJ

late EMJ (Insei)

early LMJ (Karnakura)

late LMJ (Muromachi)

NJ (Edo)

Adnominal marking

no ga

+ +

+ +

+ +

+ (+)

+

+

Subject marking in subordinate clauses and in nondeclarative main clauses

no ga

+ +

+ +

+ +

+ +

+ +

(+) +

Subject marking in declarative main clauses

no ga

+

+

Complementizer (connecting a clause to a nominal(ized) head)

no ga

+

+

Nominalizer (as pronoun and verb selected complementizer)

no ga

Conjunctional particle

no ga

(+) +

(+) +

(+) +

(+) +

+

+

+

+

+

No is normally used in the relative clause marking the second- and third-person pronouns, and it connotes deference or at least non-deprecation of the referent. In the attributive function there are two [particles]. No is used with the second- and third-person pronouns to refer to respected persons, and ga is used with the first-person pronoun and the third-person pronoun to refer to a low-ranked person, and sometimes with the second-person pronoun when that person is to be deprecated. (Shibatani 1990: 356, extracting and translating fromArte, pp. 501-3)

This passage is interesting, but it is first of all puzzling, for it is at odds with the use of ga and no in for example Esopo and other late LMJ texts, where ga is used to mark most kinds of subjects. It is difficult therefore to know what to make of this description in Arte. It suggests that this was a conspicuous feature of metalinguistic consciousness at the time, representing contemporary linguistic folklore or a rationalization of usage in EMJ or early LMJ. However, as long as there were actual personal pronouns in the language (cf. 8.8), for example, ga was used with 1st, 2nd and interrogative personal pronouns, so it is also difficult to recognize this often claimed difference in EMJ or in OJ In any case, it is difficult to link the proposed socio-linguistic differentiation of ga and no to the syntactic specialization which we find at the end of LMJ.

12.7 Honorific language

12.7

369

Honorific language

Through its history Japanese has had grammatical expression of what is very generally characterized as honorific language (keigo ~m). It is common to distinguish two independent parameters: exaltation (12. 7.1) and politeness (12.7.2). 12. 7.1

Exaltation

Exaltation characterizes the social relation between those spoken about, or between the speaker and those spoken about, and exaltation is sometimes referred to as social deixis. Two overall parameters are distinguished: (a) 'respect' (sonkei ~~), also referred to as 'esteemed subject', 'respect for subject', 'subject exaltation'~ and (b) 'humility' (kenjo ~~), which in the Japanese descriptive tradition, which focuses on the subject, may be termed 'humble subject', but which in more recent scholarship in English is referred to as 'respect for object' (or indirect object), or 'object exaltation'~ here we use the English equivalents of the Japanese terms, respect and humility, for these two categories. Of the two categories, respect (marking of subject as exalted) has always been more developed than humility. Often a number of levels of respect are set up. This may be exemplified by Rodrigues's list in Arte (p. 61) of socially differentiated ways of expressing a command, from the least to the most respectful, see (45). The 'give!' forms employ a number of morphological formants, while the forms for 'come!' are highly suppletive, using verbs that are also used as existential verbs and as polite forms (12.7.2). (45)

4

a. 'give!' aguei, agueyo aguesaxime aguesai aguesaxemaxe aguerarei voagueare voaguearo aguesaxerarei voague nasarei voague nasare6

(agei,

ageyo~ IMP)

(age-sasime~

-sasim-, IMP)

(agesai~ RESP.IMP)4 (age-sase-mase~

CAVS-mas-, IMP)

(age-rarei~ PASS.IMP) (o-age-are~ o-VERB-ar-, IMP) (o-age-arXJ~ o-VERB-ar-, VOL)

(age-sase-rarei~ CAUS-PASS.IMP) (o-age-nasarei~ o-VERB-nasar-, IMP) (o-age-nasaryoo~ o-VERB-nasar-, VOL)

A 'slightly respectful' imperative was in late LMJ and early NJ expressed by -sai (the irregular imperative of a respect auxiliary -(i)sar-) attached to regular vowel stem verbs, or -i attached to the a- stem of consonant base verbs, e.g. yomai 'read!', or irai under 'come!' in (45b).

370

12 LMJ grammar

b. 'come!' coi irai voriare vogiare gozare gozari5 voidenasarei voidenasare6

(koi; I!v.IP) (irai; ir- 'enter', RESP.IMP) (oryare; oryar- 'be', I!v.IP) (odyare; odyar- 'be', IMP) (gozare; gozar- 'be', IMP) (gozaraa; gozar- 'be', VOL) (oide-nasarei; oide 'go out' nasar- 'do.RESP', I!v.IP) (oide-nasaryoo; oide 'go out' nasar- 'do.RESP', VOL)

In this way, it is in most descriptions and in metalinguistic consciousness, as reflected in the Japanese and English terms, the social aspects of this grammatical marking and the social norms it refers to which attract the most attention. However, especially in the frequent absence of lexical or pronominal expression of subject, object or indirect object, the basic system of grammatical marking of respect and humility to some extent serves to identify those roles vis-a-vis discourse participants, as in cNJ o-yomi-ni nar- '(someone esteemed reads;) you read', or o-yomi-se- '(someone humble reads;) I read'. The system of exaltation thus grammatically has a similar function to inflection for grammatical 2nd (respect) or 1st (humility) person. 12. 7.1.1 Noun exaltation Exaltation mostly concerns clause predicates, as will be discussed immediately below, but in addition nouns and nominal words, including verbal or adjectival infinitives, may be exalted by prefixes. The primary exalting prefix is OJ miwhich is usually glossed 'honorific'. It is reflected in lexicalized forms such as miya 'palace' (ya 'house'), mikadwo 'palace; emperor' (kadwo 'gate (to palace)'), mikwo 'prince, princess, shrine maiden' (kwo 'child'). Already in OJ, the combination of opo- 'great' with mi- was conventionalized as expressing a higher degree of honorification than simple mi-, often referring to the emperor, e.g. opomi-ki 'sake for the emperor' (ki 'sake') or opomi-koto 'imperial order' (koto 'word, speech'). In the course ofHMJ, OJ opomi- was reduced to oon- (mid-EMJ) and on- (late EMJ), which replaced mi- as the general honorific noun prefix, and eventually o- which since LMJ has been the main honorific prefix. It is used widely today in a variety of functions, including formation of the 'honorific infinitive' (oyomi 'read' <= yom-; ohayoo 'early' <= haya-), with nouns as a kind of 2nd person prefix ( o-namae 'your name'), or for simple politeness (o-kane 'money' is more polite than kane), often referred to as beautification, or fully lexicalized (e.g., onaka 'stomach'). Also in the course ofEMJ, the SJ prefix go- (1i'll EMC 'imperial' *IJi:ah) gained currency, first of all with SJ words; note that 1i'll in Chinese is not used as a general honorific, but means 'imperial'.

12.7 Honorific language

371

12. 7.1.2 Predicate exaltation Overall, exaltation on predicates has been expressed in three main ways, namely by (a) auxiliaries, e.g. OJ -(a)s- RESP, -(a)sime- CAUS (enhancing a following respect expression)~ EMJ!LMJ -rare- PASS (used for respect), -saseCAUS (enhancing a following respect expression), (b) auxiliary verbs, e.g. -tamap- RESP, -mawir- HUM, -tatematur- HUM, and (c) lexical suppletion, e.g. imas- 'be, come, go, RESP', kakur- 'die' ('hide oneself'), tukapematur- 'serve. HUM', mawos- 'say.HUM, speak.HUM', mawir- 'enter.HUM', and many, many others. As mentioned, the expression of humility has always been less developed than respect, and this is also reflected in the fact that humility has never been expressed by auxiliaries, only by auxiliary verbs and suppletive lexical verbs. The inventories of forms used in these functions have changed over time, as have rules for their combinations. However, with a few exceptions the basic system of respect and humility has not changed much between OJ and cNJ. In Old Japanese, as opposed to later stages, 'respect for subject' could be used by the speaker about himself, e.g. wa ga tata-s-ere-ba I GEN stand-RESP-STAT.PROV 'as I stand' (KK 2), spoken by the god Yachihoko no kami no mikoto. This is referred to as 'absolute [as opposed to relative] respect' and disappeared in the transition to EMJ. Exaltation is ubiquitous in the literary prose texts from EMJ and early LMJ. Not only were multiple, mutually reenforcing combinations of respect expressions used, such as -saserare-tamap- CAUS-PASS-RESP, expressing a high degree of respect, but humility and respect was often expressed on one verb, usually expressing the respect of the speaker (writer) towards the subject of the verb, but expressing the humility of the subject towards the (indirect) object, for example in the combination -tatematuri-tamap- HUM-RESP. The combination of respect and humility went out of use in late LMJ and is no longer part of the system. The varied and frequent use of exalting expressions is so prominent in the literary texts from EMJ and early LMJ, especially the monogatari, that it is a distinctive feature of the appearance and image of these texts. It is, however, not likely that exalting expressions were as widely used in common language as they were in the refined and elegant language of the court ladies and nobles. Through LMJ, the use of exalting expressions in the texts declined somewhat, probably reflecting more everyday and less elegant language use, rather than structural changes. In cNJ, exalting language has been regularized to mainly morphological expression, with a small number of auxiliaries and auxiliary verbs and far fewer suppletive verbs than in EMJ or LMJ. It should finally be mentioned that despite the prominence of the use of exalting expressions in metalinguistic consciousness and the importance assigned to it in normative discourse about Japanese, a number of Japanese dialects do not use exalting language.

372

12 LMJ grammar

12. 7.2

Politeness

As opposed to exaltation, politeness (teineigo T·~ 'polite language') is a matter of style or register and expresses relations between speaker and hearer without reference to the content of the discourse. Politeness is a prominent feature of cNJ where polite style is expressed by the auxiliary -(i)mas- or the polite copula des-, but, as with exaltation, there are NJ dialects which do not have polite style. The grammatical expression of polite style is a relatively late addition to the language: OJ did not have it at all, and while there are incipient uses in EMJ, it was not thoroughly established until early LMJ. Polite style originates in exalting expressions through a process of shifting the target of respect or humility from the subject of a sentence to the hearer (respect) or speaker (humility), eventually interpreted as a characterization of the speech situation, or the relationship between speaker and hearer. Thus, all material used to express polite style originates in respectful or humble forms, and, as with exalting auxiliaries and auxiliary verbs, the development of the individual polite style markers is often complex. Around the middle of EMJ, paber-, which until then was used as a humble verb 'be in attendance' or a humble equivalent of the existential verb ar-, started being used as a polite equivalent of ar- in its various uses. Thus, the expression of politeness was at first limited to contexts where ar- was used (existential sentences, extended regular and adjectival copula forms, stative constructions~ cf. 12.4). However, in early LMJ, paber- was in this function replaced by sauraw-, which in addition to being used as a suppletive polite equivalent of ar- came to be used as an auxiliary verb, that is, attaching to the infinitive of verbs to express polite style, e.g. mi-saurawabaya 'see-POL. OPT~ if only I could see', which is an early example from the He ike monogatari (early thirteenth century). This development made possible the expression of polite style with all predicate types and established politeness as a grammatical category in the language. Sauraw- comes from early EMJ saburap- which is said to be a changed form of OJ samorap- 'serve, be in attendance', also reflected in the word samurai. During LMJ -sauraw- was used as the general polite marker~ it changed to -sa:vaw- by regular sound change (11.5), but also gave a number of other more reduced shapes with irregular inflected forms, such as saa(-) which was both used on its own as the nonpast form and as a stem for attaching other morphemes: saata past, saanu negative. This divorced its polite-style marking function from its use a lexical verb. Thus, S:J:Jr::xJ and its different variants might best be understood as an auxiliary, rather than as an auxiliary verb, but it should be noted that they continued to function as polite variants of ar- in all its functions, including combining with the copula gerund de to form polite copula forms, e.g. de s:J:Jr:J:J or de s:J:J. The cNJ polite copula des- is by some

12.7 Honorific language

373

scholars thought to descend from de s::x:J Sora-bun is the name of a formal style of writing, which developed during NJ, which is characterized by extensive use of NJ sooroo to conclude sentences. LMJ had several more polite style synonyms of ar-: odyar- (< o-ide-ar-; ide- 'go out') and oryar- ( < o-iri-ar-; ir- 'enter') both originate in respect verbs for 'come, go' and further 'exist'. They came to be used widely in late LMJ as polite style verbs (especially as ar- equivalents), but disappeared during NJ. Also gozar-, which is still used widely in cNJ as a suppletive (super-)polite existential verb, originated in LMJ as a respectful equivalent of ar-. It was originally a compound of goza and ar-, and through LMJ had the alternative shape goza ar- and the negative form goza na-. Goza arose in early LMJ as a SJ reading of~~ which had been used as a kun-writing of the respectful existential verbs owas- and owasimas- (see 9.2.4). At the end ofLMJ, gozarwas the more frequent shape and the verb was in frequent use. Finally, in the second half of LMJ, -(i)saaraa was replaced as the general marker of polite style by the auxiliary -{i)marase-, which is the source of the present-day polite style auxiliary -{i)mas-, e.g. from Esopo: i-marasuru existPOL.NONPST 'is', osie-marasyoozu teach-POL.INT '(I) will teach'. As opposed to the other polite style verbs and auxiliaries mentioned above, -{i)marasedoes not originate in a suppletive respectful or humble existential verb. Its OJ source is the humble verb mawir- 'come.HUM, go.HUM' (which itself is reconstructible as *maw+ ir- 'humble prefix+ enter'), through lexicalization of the causative mawira-se- to a suppletive humble verb mawirase- 'give.HUM' >(by regular sound change, 7.3.2.3) mairase-, which came to be used as a humble auxiliary verb -mairase-, which in tum in late LMJ was further reduced phonologically and became an auxiliary -{i)marase- which shifted from humble to polite, eventually being reduced even further to its present-day shape -(i)mas-; this is thought to have taken place early in NJ, but there are examples from LMJ which suggest that it may have happened already in late LMJ and that the two shapes -(i)marase- and -(i)mase- coexisted for some time during LMJ, but that the latter was only sporadically reflected in writing. REFERENCES General: Imaizumi 1971, Iwai 1971, 1973, Rodrigues 1604-8. Shomono: Yuzawa 1929/1970. Feiqe: Yuzawa 1928. Esopo: Otsuka and Kita 1999 (includes a classified index to Esopo). Conclladn merger: Yamauchi 2003. Existential verbs: Kinsui 2006. No as subject marker: Yanagida and Whitman 2009. Adnominal clauses and nominalizations: Frellesvig and Whitman forthcoming, Kinsui 1995, Shibatani 2009, Wrona 2008: 237ff., forthcoming.

Part IV

Modem Japanese

13

Varieties of Modern Japanese

13.1

The Edo period: Linguistic diversity and common language

In the Edo period, geographical and social mobility was very low, and most people stayed all their life in the same place within provinces which were in effect isolated feudal states. This situation resulted in a large number of fairly small and self-contained speech communities. The linguistic diversity with lack of mutual intelligibility between many dialects within Edo-period Japan is famous and is well illustrated by the following extract from Furukawa Kosh6ken's Toyil zakki (Jii[Jbf~~e 'Notes from a journey to the east'), an account of a journey in 1788 to northern Honshu and Hokkaido, accompanying inspectors from the central government inEdo, where he describes experiences with the local dialects they encountered (cited from Shibata 1998/1965: 184). Of the dialect in Tajima (in present-day Fukushima prefecture), Furukawa writes: Both sides can Wlderstand only half of the other's language. There was nothing one could do but laugh about it even at the inns. When we asked for cha-zuke 'tea on boiled rice' they broughtyu-zuke 'hot water with boiled rice' instead, and we would go into the inn kitchen each time and make it ourselves.

And about the dialect in Nanbu (in present-day lwate prefecture): The language of both men and women was gibberish, with only two or three words out often being comprehensible. The local lords usually give each inspector, in addition to his guides, two or three people well-informed about the castle town for the duration of his stay. The lord of Morioka Castle gave us two interpreters in addition because the speech of this part of the Nanbu domain is notoriously incomprehensible. But even the interpreters often could not Wlderstand what people say here.

13.1.1

Cammon language

Although the centre of political power shifted away from Kyoto with the establishment of Edo as the de facto capital by the Tokugawa shogunate at the beginning of the seventeenth century, Kyoto retained its position of the 377

378

13 Varieties of Modem Japanese

prestige centre of high culture through the Edo period, and its language the position of prestige language. While not having the standing of a 'standard' language, a form of Kyoto Japanese was in use throughout Japan and functioned at the beginning of the Edo period as a common language among the political and military elites, high ranking officials, clergy, etc., in addition to the many dialects of great diversity all over Japan used by local people.

13.1.2

The language ofEdo

Edo was a small castle town at the beginning of the seventeenth century. Its growth since then has taken place mainly by immigration. Through the Edo period, the system of 'alternate attendance' (sankin-kotai) brought a constant flow of local feudal lords and their entourages to Edo, which also attracted significant numbers of unemployed warriors and which had a growing merchant and artisan population. Life in Edo was characterized by frequent and diverse social intercourse between the inhabitants and visitors and by a lively popular culture and entertainment industry. Through the second half of the Edo period, Osaka developed similarly to Edo as a socially diverse centre for commerce and trade, also with a rich popular culture. These two settings thus made for dynamic linguistic environments which were quite different from the rural settings in which most people lived, or from the socially segregated life in Kyoto, and the language of both Edo and Osaka developed accordingly. The language of Edo/Tokyo took form during the Edo period, but the process of its formation is difficult to trace in detail. It is clear, however, that Edo Japanese is no direct continuation of any particular local dialect of Japanese, but came into being and developed through close contact between speakers of different varieties in an urban setting. In its formative phase, the language ofEdo involved a significant component of the Kyoto-based common language, which was influenced by various dialects and by its use in the urban setting of Edo. The colloquial local language of Edo which developed through the period and which is attested particularly from the middle of the eighteenth century is certainly characterized by a wealth of occasional, fashionable, jocular or specialized vocabulary, by formal styles employing different honorifics, as well as by informal speech styles with noticeable phonological reductions~ a stereotypical example is arinsu which was a polite auxiliary verb reduced from arimasu, used mainly by prostitutes in the red-light district of Yoshiwara. However, these are in many cases ephemeral features of everchanging and renewing language usage, typical of a growing urban setting, and many of the special features which may be found in the colloquial language of Edo have not survived into the contemporary language. Alongside the developing urban popular idiom, Edo Japanese, a form of the Kyoto-based common language continued to be used in Edo by the upper

13.2 The Meiji period

379

classes and towards the end of the period also by the emerging educated middle class. This is the variety which eventually came to form the basis for standard cNJ. Although it was influenced by the local Edo language, it changed slowly, and it maintained its contiguity with the language in Kyoto. Differences between the versions of the common language used by the upper classes in Kyoto and in Edo developed, but the users of the common language continued to make up a socio-economically constituted, but geographically discontinuous, large speech community, with some internal variation, which played an important role through the NJ period, at least until the beginning of the twentieth century. 13.2

The Meiji period: Unification and standardization

Meiji was a period of unification in political terms: the creation or building of a unified, modem nation state. This unification effort also affected language, in two main ways: one was the unification, or alignment, of the spoken (i.e., contemporary) and the written language (13.2.1); and the other was the unification of the different varieties of Japanese spoken through the country as, or under, one national language, that is to say, the creation of a national, standard language (13.2.2). 13.2.1

Genbun'itchi

As mentioned above, Japanese was throughout the Edo period written in a variety of ways which had in common that they reflected the contemporary language very indirectly or not much at all, mostly being very convoluted and involving some form of the classical written language. In addition to the fact that there was no established way of representing the contemporary language in writing, the effort and time required to become literate were considerable. In the second half of the nineteenth century, with the opening of Japan and establishment of commercial and diplomatic contact with European countries and the US, an appreciation grew among educators, reformers and modernizers that the way in which Japanese was then written hindered literacy, education and, more generally, modernization. From the 1860s through to the first decade of the twentieth century, there were fierce debates over whether, and how, to write the contemporary language, with lines sharply drawn between proponents and opponents of reform. From the mid 1880s authors who wanted to write realistic novels joined the movement, or campaign, for vernacularization of the written language, which came to be known as genbun 'itchi ('unification of speech and writing', Jt-jl:). The serial publication in 1887-9 of Japan's first modem novel Ukigumo by Futabatei Shimei (1864-1909), which was written in the vernacular, was an important landmark which gave the campaign

s

380

13 Varieties of Modem Japanese

additional momentum. From then onwards more literature written in the vernacular became published, and also newspaper editorials gradually adopted a vernacular-based written form. The first school textbooks with large proportions of vernacular language were issued by the Ministry of Education in 1903-4. Although a form of the classical written language was still used in many newspaper articles until the 1920s, and until the end ofWWII in official documents and government decrees, and even in some academic publications (for example Yamada Yoshio's grammars of OJ and EMJ from 1913), the reformers had for all practical purposes won by the mid 1910s, with the vernacular adopted for education, literature and much public written discourse. As mentioned in 6.1.3, reform of the kana orthography with its etymological spelling principle (for example spelling /kyoo/ 'today' as tt b "ke.fu"), did not take place until after WWII.

13.2.2

A national language; standardization and dialect eradication

Part of the modernization effort and the establishment of a unified nation-state was also the notion of a unifying national language. Especially after the publication of an essay collection entitled Kokugo no tame 'For the national language' by Ueda Kazutoshi in 1895, in which he drafted in the SJ word kokugo (00~ 'country-language') for this notion of Japanese as national language, the word kokugo has become widely used to refer to the Japanese language, and the word still has strong emotional value for some. For example, the renaming in 2001 of the 'Society for Japanese Linguistics' from Kokugo gakkai to Nihongo gakkai caused rifts in Japanese academia which have yet to be healed. Kokugo is an exclusive, somewhat nationalistic term, for it can only be used to refer to the Japanese language by and for the Japanese. However, both the notion of national language and the term for it were in the twentieth century exported to Korea, where 00~ kug-6 means 'the Korean language'. Both the notion of kokugo and the practical purposes of the vernacularization of the written language (spreading literacy and education) pointed in the direction of establishing a standard language, that is to say, privileging one variety over others for use in public life, including education, and as an emblem of the nation. Thus, in 1901, the Ministry of Education decreed that the Japanese language taught in schools should be that of Tokyo, whereby was meant not the language of the common people of the downtown area, but the language of the middle and upper classes of the Yamate area, in other words the descendant of the Edo-inft.uenced variety of the common language (13.1). One example from phonology of a well-attested Edo feature which did not pass into the standard language is the monophthongization of /ail as [ e:]. This is attested from the early eighteenth century and is still today a colloquial feature of Tokyo speech, but is not sanctioned in the standard language.

13.3 Sources

381

In 1902, the Kokugo-chOsa-iinkai (00m~1i~ffit~ 'National Language Research Council') was set up, charged with surveying the state of the national language and making recommendations, amongst other things, for establishing a standard language. The Kokugo-chOsa-iinkai was dissolved as a body in 1913, but its recommendations regarding a standard written language were published inKogoho (omrt 'Grammar of the vernacular', 1916) andKogoho bekki ( omrtZIJ~C. 'Supplement to Grammar of the vernacular', 1917). It was in these volumes that the normative grammar of standard Japanese as we know it today was set out and where it was clearly said that standard Japanese was based on the speech of the educated middle and upper classes of Tokyo. The chief editor for these two volumes was Otsuki Fumihiko (1847-1928), so if any individual is to be credited with the creation, or at least codification, of modem standard Japanese, it is he. While standardization and vemacularization of the written language has had many positive effects in terms of spread of literacy and education and increased popular participation in public life and eventually also in the political process, the ft.ipside was a directed and quite successful effort to eradicate dialects, especially through the first half of the twentieth century, leading to a marked loss of linguistic diversity in Japan which is regrettable, at least to linguists and ethnographers. In the pre-war years, users of dialect in schools were subjected to public ridicule and even punishment. For example, in some schools those overheard using the local dialect rather than the standard language were forced to wear hOgenfuda 'dialect tags'. It was only after WWII that the active discouragement of use of dialects was relaxed, but even today many people are ashamed of speaking their local dialect in public. Recently, an Osaka-based variety of Kansai Japanese is getting wider public exposure, especially in popular culture, and is even thought fashionable and imitated by some young people. Even so, many speakers of Tokyo Japanese will profess- somewhat disingenuously- an inability to understand Kansai Japanese.

13.3

Sources

Present-day Japanese is observable and well described. We have a large amount of material from the Meiji period onwards (second half of the nineteenth century), both in Japanese script and in alphabet writing, which shows that the language has not changed much since then. Before Meiji, throughout the Edo period, Japanese was written in a variety of ways which had in common that they reflected the contemporary language very indirectly or not much at all, mostly being very convoluted and involving some form of the classical written language. These ways of writing ranged from (hentai) kanbun over various forms of Classical Japanese to highly formalized versions of post-Classical Japanese, which was, however, still heavily influenced by the

382

13 Varieties of Modem Japanese

classical written language. There is, however, also a large body of popular fiction and drama in which the contemporary vernacular was used, especially in renditions of spoken language, namely in lines in plays and in dialogue parts of novels. Until the middle of the period, the language reflected in these sources was mostly that of Kyoto, but from the second half of NJ, literature which reflects the emerging urban idiom of Edo was increasingly published. The Edo sources are far too numerous to name. In addition to kabuki and puppet play scripts, the drama material includes kyogen, which are lively short comedies or sketches, originally performed in between noh plays as a kind of comic relief. Kyogen plays gained independent popularity from the Muromachi period and at that time they were performed in the contemporary spoken language. However, over time a formulaic style of language evolved which is reflected in the great majority of the surviving texts which date from the mid seventeenth century onwards. As the plays were originally handed down in performing traditions, the NJ kyogen texts in fact reflect many features of the language of the Muromachi period. 13. 3.1

Material in alphabet writing

We have a wealth of material in alphabet writing from the second half of the nineteenth century which is very close to the contemporary language. Two earlier decriptions of Japanese are, however, worth mentioning. The first is a seventeen-page article, published in 1792 in Uppsala, 'Observationes in Linguam Japonicam' by the Swedish botanist Carl Peter Thunberg who stayed in the Dutch trading post in Dejima off Nagasaki for around a year and a half during 1775-6. In it Thunberg gives a brief description of the Japanese language. The language he describes seems to be common educated Japanese, including, however, identifiable Kyushu dialectal features. Thunberg is clearly no linguist, but his little piece contains valuable information on late eighteenth-century Japanese. Phillipp Franz Balthasar von Siebold was a German physician who taught medicine in Japan from 1823 to 1829 (when he was expelled from Japan accused of being a spy because he had collected maps). His seventy-page long Epitome linguae japonicae was written in Nagasaki in 1824 and, again, has valuable information about Japanese from the early part of the nineteenth century.

13.4

From Late Middle Japanese to standard contemporary Modern Japanese

As outlined above, the colloquial urban idiom ofEdo and Tokyo, Edo Japanese, is not a direct descendant of the language of Kyoto OJ, EMJ or LMJ reflected in the written sources, because of its multiple sources and influences (although

13.4 From LMJ to standard cNJ

383

its initial main component was the common language). However, the situation is different for the version of the common language used in Edo!fokyo upon which the standard language came to be based. Although it was undoubtedly influenced by Edo-Japanese, it represents a fairly unbroken tradition of the Kyoto-based common language. Thus, comparing the language offor example Esopo, which is from the very end of LMJ and which reflects the common language of Kyoto, with cNJ, what is really striking is how relatively speaking few and small the structural differences are. In fact, there are very few features of cNJ which are not straightforwardly derivable from LMJ as reflected in the sources from the end of the period. In the following the main phonological and morphological differences between LMJ and cNJ will be outlined. In a few cases differences between LMJ and cNJ seem to reflect influence from eastern dialects (mediated through Edo Japanese). However, most phonological and morphological changes took place through the common language and are reflected in both Kyoto and Tokyo cNJ, demonstrating the coherence and influence of the common language speech community at least until the beginning of the twentieth century. These changes, outlined in Chapters 14 and 15, are describable as simple linguistic changes between LMJ and cNJ. Features which are now part of the standard language, but were not used in Kyoto LMJ or NJ, and which are thought to reflect eastern dialect influence, will be discussed below (in Chapter 16). REFERENCES General: Iwai 1974, Matsumura, 1998, Sugimoto 1988, Yuzawa 1957, 1970. Women's language: Endo 2006, Sugimoto 1985. Unification, standardization and dialect eradication: Ramsey 1991, Shibata 1998/1965, Twine 1991. Western sources: Sugimoto 1989.

14

Phonology

Table 14.1 gives the by now very conservative sound inventory of free moras in cNJ. The phonemic representation in Table 14.1 is used in the kunreishiki (~}114';-:lt 'official directive system') system of transcription, which in a modified version is used by most linguists, whereas the more broadly used Hepburn system reflects allophonic variation (see 14.5) and uses letter combinations which appeal to English readers: /si/ shi, /til chi, ltul tsu, lhul fu, lzil ji, /sya/ sha, /syu/ shu, /syo/ sho, ltyal cha, ltyul chu, ltyol cho, lzyal ja, lzyul ju, lzyol jo. Some of this originally allophonic variation has since become phonemic in recent cNJ (see 14.6) and that is the main reason we have chosen to use the modified Hepburn system in transcription of cNJ in this book. During NJ a few sound changes took place. The precise dating of them is not easy for there are no extensive materials in alphabet writing until the end of the Edo period. The sound changes which took place between the end of LMJ and conservative cNJ are outlined in the following. All the distinctions which were lost in the course of these changes are preserved dialectally.

14.1

Merger of /d/ and /z/ before /i, ul

Through the latter part of the LMJ period It, d/ became phonetically assibilated before the two high vowels, li, u/, as reflected in the Christian sources (11.6). Early in the NJ period, this further led to the merger of /d/ and /z/ before /i, u/~ the outcome of the merger is considered to be /z/, which however continues to have affricated realizations: (1)

/d/ > /zl I li, u/ /du/ [d~] 1 /zul [zu] > /di/ [d3i] 1 /zi/ [3i] >

/zu/ [dzu- zu] /zi/ [d3i]

midu 'water' 1 mizu 'not seeing~ doesn't see' > midika- 'close' 1 mizika- 'short'>

mizu mizika-

This change is often thought of as a merger between the moras di 1 zi and du 1 zu, and in Japanese it is referred to in terms of the kana letters representing 384

14.2

Me~ger

of /oo/ and fool;/':)':)/> fool

385

Table 14.1 Free moras in conservative cNJ a i u e 0

ka ki ku ke ko

sa si

ta ti tu te to

ga gi gu ge go

za zi zu ze zo

da

kya kyu kyo

sya syu syo

tya tyu tyo

gya gyu gyo

zya zyu zyo

su

se so

na ni nu ne no

ha hi hu he ho

pa pi pu pe po

rna mi mu me mo

pya pyu pyo

mya myu myo

ya

ra ri

yu

ru

yo

re ro

wa

ba bi bu be bo

de do nya nyu nyo

hya hyu hyo

rya ryu

ryo

bya byu byo

these moras, the yotsu-gana 'four kana', viz. t:, i:)f--:5. The Christian materials keep the four moras distinct, noting them asji lzil, gi /dil, zu /zul and zzu /du/, respectively, but the merger seems to have taken place in the first half of the seventeenth century. As mentioned in 6.1.3 above, spelling conventions still today keep t versus 16, and f versus ---5 distinct in compound contexts, for example spelling min_ka- 'close, familiar' as li- i6 lJ\ reflecting its origin in a compound mi 'body'+ ti_ka- 'close', or kana-zukai 'kana-usage, spelling' as tJ-,.ft.---5"/J-,.v\ cf. tsukai 'usage'.

14.2

Merger of /oo/ and /oo/; fool > /oo/

Also early in the NJ period, the distinction between !:J:J! (
I':J':JI > fool

t:J:Jto- 'precious'

-=F

imooto 'little sister' > tooto-, imooto

This change affected in particular the shapes of a number of SJ vocabulary items (see 11.5.2), and the morphophonological rules applying to inflected adjective and verb forms ending in -u (12.3.3), which changed so that both /au/ and lou/ gave /oo/, see (3). See below (15.1.1) about the volitional whose

14 NJ phonology

386

formation changed so these rules became irrelevant, and see 14.2.1 about the nonpast of -w- base verbs. (3)

LMJ /au/=> !:J:J! lou/=> fool

(4) a.

b.

14.2.1

> >

NJ /au/=> fool lou/=> fool

Verbal u- onbin stem kaw- 'buy' kau-te => kau-de => kam- 'bite' yow- 'get drunk' you-te => yam- 'read' you-de=>

koote koode yoote yoode

Adjectival copula infinitive -u taka- 'tall' taka-u => kuro- 'black' kuro-u =>

takoo kuroo

/Vu/ diphthongs

In the course of the LMJ contractions of /Vu/ sequences to long vowels, /Vu/ diphthongs were eliminated at the phonemic surface level by sound changes (cf. 11.5) and by morphophonological rules such as those discussed immediately above which contracted sequences of IV + ul which arose in the formation of some inflected forms (12.3.3). However, in the course ofNJ, this rule stopped applying to the formation of the nonpast of -w- base verbs, so that in cNJ the nonpast of aw- 'meet' is au and yow- 'get drunk' is you, not oo or yoo. This, combined with the intake of loanwords such as mausu 'mouse (especially for use in medical experiments)' (from German Maus), reintroduced /Vu/ sequences in the language. Apart from the very recent relaxation of restrictions on the occurrence of /Q/ (see 14.6 below), this was the only change which affected long syllables in the NJ period.

14.3

Delabialization of Iff; Iff > /hi

The final change to affect the reflex of OJ /p/ was the change of /f/ (see 11.3) to /hi which occurred at some point during NJ. This change was primarily a phonetic change with no mergers or other immediate phonemic impact and it was therefore not reflected in general writing. Thunberg (1792) mostly writes words reflecting LMJ /f/ with 'f' (e.g. Janna 'flower'), but says that words with 'f' sometimes are pronounced ('read') with 'h' (e.g. hanna) and Siebold (1826) uses 'h' in his transliteration of the kana letters f'J:"(}J,...r-..f~, but notes that they are often pronounced with 'f'. However, by the middle of the nineteenth century, sources in alphabet writing generally write 'h', except before

14.3 Delabialization of Iff; Iff > /hi

387

'u' where it is 'f', reflecting that !hi is pronounced [cp] before /u/. Hepburn's transcription system from the end of the nineteenth century adopted this allophonic notation which since then has dominated transcriptions in the west. Very recently, due to a large intake of loanwords from English, [cp] has become phonemic /f/ before vowels other than lui (14.6).

14.4

Loss of phonetic onglides

Two changes affected glides: one was the loss of the automatic onglides in the syllables /.o/ and I.e/, see (5). This happened earlier to /.of which is transcribed 'o' in Thunberg (1792) and Siebold (1824), whereas /.e/ is transcribedye into the Meiji period, as reflected in old spellings such as Yedo for the capital, as well as spellings in English dating from that period, e.g. yen for the Japanese currency, Yebisu for the beer brand, or ue spelled uye in for example Inouye and other personal names, especially by Japanese-Americans. (5)

LMJ /.of=> [wo]

/.e/ => [ie]

NJ > >

/.of=> [o] I.e/=> [e]

Second, the labial glide which had survived in /kwa, gwa/ in SJ words such as kwaiwa (~~!)'conversation', kwasi (Jit-1-) 'cookie', or gwannen (:lC'¥) 'first year in a dynastic period' was lost, so that these syllables merged with /ka, gal (kaiwa, kasi, gannen). Again, this change was completed late in the nineteenth century, and the glide is reflected in occasional spellings such as Kwannon (Kannon, the Goddess of Mercy) or in Kwansei Gakuin Daigaku, the name of a private university in Kobe. 14. 4.1 Loss of palatalization before lei Allophonic palatalization of consonants before /e/ was lost, probably related to the loss of the automatic glide before /.e/. However, palatalization of all consonants is retained before li/. Thus siri [Jiri] 'buttocks', but seri [seri] 'parsley' < LMJ [Jeri].

14.5

Summary of main allophonic variation

The most conspicuous allophonic variation in conservative cNJ is summarized in (6). We do not exemplify palatalization of all consonants or aspiration of /p, k/. Intervocalic nasalization of /g/ is on the way out and is for many speakers today a normative feature requiring conscious effort. As in LMJ, vowels in final position are often weakened, as are li, u! between consonants, especially, but not exclusively, if these are phonetically unvoiced.

388

14 NJ phonology

(6)

It/=>

[th] I_ /e, a, o/ [ts] /_lui [affrication]

[1j] I_ li, y/ [affrication, palatalization] /s/ =>

[s] /_ /e, a, o, ul [J] I _li, y/ [palatalization]

/hi=>

[h] /_ /e, a, o/ [
/z) =>

[z] /_ /e, a, o/ [ z, dz, dz] I_ { u} [affrication]

[<:B] I _li, y/ [affrication, palatalization]

/g/ =>

14.6

[g] I#_ [IJ] IV_V

Recent phonemic changes due to loanwords

In the course of the intake of loanwords from the beginning of the Meiji period, but particularly after WWII, some phonemic changes have taken place, which are either phonemicization of the phonetic variants outlined immediately above, and/or new combinations of existing phonemes. Thus, in recent cNJ, the following sounds which in conservative cNJ are conditioned variants are now distinctive before /e/, in addition to before /a, o, ul (7a). This was made possible by the loss of automatic palatalization before /e/. Furthermore, [t] has become phonemic before/i, ul(7b), and, marginally, [ts] is now found before othervowelsthan lui, (7c). Finally, [
(7) a.

[J]

sherf 'sherry' [Jeri:] versus seri 'parsley' [seri]

[tJ]

chen 'chain' [tJe:N] versus teinai (~~1*1) 'premises' [te:nai] jerf 'jelly' [<:Beri:] versus zero 'zero' [zero]

[cB] b.

[t]

piitf 'party' [pa:ti:] versus pachinko 'pachinko' [patj"il)ko] Turu 'Tours' [tu:ru] (city in France) versus tsilru 'tool' [tsu:ru]

c.

[ts]

tsaitogaisuto 'zeitgeist, spirit of the times' [tsaitogaisuto] kantsone 'canzone' [kantso :ne]

d.

[
fan 'fan' [<PaN] versus han- 'half' [haN]

14.6 Recent phonemic changes due to loanwords

389

Due to sound changes which took place in EMJ and LMJ, /w/ was lost before other vowels than /a/ (11. 8.1 ), but recent loanwords have re-introduced /w/ to a limited extent before li, e, of, e.g. wirusu 'virus', wesutan 'western (movie)', wokka 'vodka'. Finally, [v] is used by some educated speakers in free variation with [b] or [w] in some loanwords reflecting foreign [v]: vandaru-zoku (- bandaru-zoku) 'the Vandals (Germanic tribe)' (SJ -zoku ~ 'tribe'), venetsia- benetia 'Venice', vaimiiru- waimiiru 'Weimar'. In long syllables /Q/ was until cNJ restricted to occurring before tenues (cf. 11.1.2), but it now also occurs before mediae e.g. beddo 'bed' /beQdo/, baggu 'bag' /baQgu/, karejji 'college' /kareQzi/. Long faa! (Ia VI) seems not to have been used outside expressive forms until the analogical formation of the distal adverb ii 'that way' (corresponding to ko 'this way' and so 'that way'), which is attested in the late eighteenth century. From the Meiji period faa! is found in loanwords such as giiden 'garden'. REFERENCES Mabuchi 1971, Martin 1987, Vance 1987, Wenck 1959.

15

Grammar

15.1

Verbs

The main inflected standard cNJ verb forms are shown in Table 15.1. In terms of morphological categories, the main differences between the LMJ (12.1.1) and NJ paradigms are the loss of the tensed non-finite forms (12.1.3.3) and the loss of the intentional (12.1.3). The new representative reflects the EMJ stative auxiliary -(i)tar-; there are examples of this use of -(i)tari in LMJ, but it does not become established as an inflected form until NJ. The past conjectural is no longer a productive form in cNJ, but is used in old-fashioned writing and in some dialects. See below regarding the shape of the nonpast (15.1.2.1), the past, gerund, conditional and representative (16.1), the volitional (15.1.1), and the imperative (16.4). 15.1.1

Formation ofthe volitional

In LMJ the volitional was formed by -u, attaching to the a- stem of consonant base verbs and the basic stem of vowel base verbs, e.g. kaka-u => kab:J, ake-u => akyoo (12.3.3). In NJ the shape and formation of the volitional changed to give the NJ forms shown in (1):

(1)

390

kak- 'write' kaw- 'buy'

LMJ kaka-u => kab:J kawa-u => ka:J:J

NJ kakoo kaoo

ake- 'open' ne- 'sleep'

ake-u => akyoo ne-u => nyoo

akeyoo neyoo

oki- 'arise' mi- 'see'

oki-u => okyuu mi-u => myuu

okiyoo miyoo

ko- 'come' se- 'do'

ko-u => koo se-u => syoo

koyoo shiyoo

15.1 Verbs

391

Table 15.1 Standard cNJ inflected verb forms

Basic stem

QD

QD

LB

leak-

kaw-

ake-

kaku kaita kakoo Ieake (kaitaroo)

kau katta kaoo kae (kattaroo)

akeru aketa akeyoo akero (aketaroo)

kaki kaite kaitara kaitari kaitewa kakeba kaitemo

kai katta kattara kattari kattewa kaeba kattemo

ake akete aketara aketari aketewa akereba aketemo

Finite

Nonpast Past Volitional Imperative (Past conjectural) Non-finite

Infinitive Gerund Conditional-! Representative Conditional-2 Provisional Concessive

Apart from the merger of kd and fool (14.2), which also affected the shape of the volitional of the consonant base verbs, e.g. kakaa> kakoo, the only change was in the shape of the vowel base verb volitional. This change is probably related to the levelling and greater stem transparency in the vowel base verb conjugations. Whether the volitional forms are analysed as (a) kak + yoo => kakoo and ake + yoo => akeyoo, with deletion of the initial glide in the suffix with the consonant verbs, or (b) kak + oo => kakoo and ake + oo => akeyoo, with automatic glide insertion with the vowel base verbs, is mainly a theoretical question. In any case, the new formation resulted in greater uniformity and transparency in them orphological composition of this form across the conjugation classes, with suffixation of a ftective (-yoo or -oo) directly to the basic stem of verbs from all conjugation classes (except the irregular s-irr class). 15.1.2

Verb classes

The number of conjugational verb classes was reduced during the NJ period, to give the five classes of cNJ. 15.1.2.1 Levelling ofvowel base verb conjugations; merger ofmonograde and bigrade verbs From OJ through LMJ most vowel stem verbs belonged to the bigrade conjugations (LB and UB), which had stem alternations between a basic stem, e.g. ake-, and a derived form, aku (<= ake + u), which in OJ and EMJ was the

392

15 NJ granunar

conclusive fonn also functioning as a stem for formation of further forms (3.4.1.3). With the concl/adn merger and the loss of the exclamatory fonn in LMJ (12.1.2), the derived fonn was no longer an independent word fonn, but only a derived stem which was used in the fonnation of the nonpast fonn (e.g. akuru <= aku- + ru). The monograde vowel stem verbs, UM and LM, had no such alternation between a basic and a derived stem, but used the basic stem in fonnation of all fonns. In EMJ short (one-mora) UB verbs shifted to monograde conjugation, so that short i-stem verbs belonged to UM, and stems of more than two moras belonged to UB (8.1.2), but other than that the difference between bigrade and monograde vowel stem conjugation remained through LMJ. However, in the course of the first half ofNJ, bigrade conjugation was lost in most dialects (and is today retained only in a very small number of dialects) and all LB verbs became LM, and UB verbs became UM. As part of this change the shape of the nonpast changed so that it too came to be formed on the basic stem, see (2a) below. With the changes in the fonnation of the volitional (15.1.1), the levelling of the vowel base conjugations resulted in greater transparency in the morphophonology of the vowel base verbs, with direct manifestation of the basic stem in all inflected fonns. The change of the LB verbs to LM took place first with short verbs and at the end ofLMJ a few short -e base verbs were in the process of shifting and had variant nonpast fonns, e.g. fe- 'pass': foru- feru.

15.1.2.2 Merger ofn-irregular and quadrigrade verbs The (irregular) n- base verb sin- lost its irregular conjugation and became a regular QD verb. As with the levelling of the bigrade verbs to monograde, this change mainly affected the formation of the nonpast, (2b): a.

ake- 'open' ne- 'sleep' oki- 'arise'

LMJ akuru nuru okuru

NJ akeru neru okiru

b.

sin- 'die'

sinuru

sinu

(2)

15.1.3

Summary ofverbal conjugation classes

The general changes among verbal conjugation classes from OJ to NJ are summarized in Table 15.2. 1 The main change between OJ andEMJ (see 8.1.2) 1

In Japanese school grammar, the regular consonant base verbs in cNJ are called godan (' quinquigrade') instead of yodan, reflecting that the katsuyokei system posits an extra stem, ending in -o, whose only function is to be used to form the volitional by attaching -o. Thus, in the katsuyokei system kale- 'write' has the suiryokei ('conjectural stem') kako- to which -o is attached to give the volitional kakoo. The sole motivation for this analysis is to fit the formation of all forms into the katsuyokei system.

393

15.2 Adjectives and copula Table 15.2 Changes among verbal conjugation classed from OJ to NJ

QD r-irr n-irr

LM LB UM UB k-irr s-irr

OJ

EMJ

LMJ

NJ

CVCarsin-

CVCarsin-

CVC-; ar-

CVC-, ar-, sin-, leer-

sin-

IceCe-, CVCe-

Ce-, CVCe-

CVCe-, Ce-

kweCe-, CVCe-

Ci-,CwiCVCwi-, Cwi-

CiCVCi-

CiCVCi-

Ci-, CVCi-

kose-

kose-

kose-

kose-

was the emergence of the LM class with EMJ kwe- > LMJ ke- as its only member, and the migration of the short (one-mora) UB verbs to UM, so that in EMJ!LMJ short -i base verbs were UM and more than two -mora -i base verbs were UB. In the course of LMJ, the r-irr class merged into the QD verbs as a result of the conclladn merger (12.1.2). Already at the end of LMJ a few short LB verbs were in the process of migrating to LM, but in NJ the bigrade conjugation was lost and the UB verbs merged with the UM verbs to become UM, and the LB verbs became LM (15.1.2.1); however, the sole EMJ!LMJ LM verb ke- 'kick' migrated to QD and is now ker-. In addition to these general changes, a number of individual verbs have migrated between conjugation classes through the history of the language.

15.2

Adjectives and copula

The paradigms of both the adjectival and the regular copula show a simplification compared with LMJ. The main cNJ forms of the adjectival copula are shown in Table 15.3. Comparing the LMJ (12.2.1) and NJ paradigms, the main differences are the establishment of an inflected past tense for the adjectival copula as with the verbs and the shift from expressing negation by an auxiliary -n- (-karan-) to an analytic construction with na- (or polite arimasen) following the infinitive (-ku na-, -ku arimasen); see further 16.3. The LMJ concessive, -keredomo was lost (replaced by -kutemo), as was the LMJ conditional, -kuwa ~ -kunba (replaced by -kattara, which reflects the secondary adjective conjugation, supplemented with -kutewa). Both -kutemo and -kutewa conform to the same pattern as the verbs of using the gerund as a stem for further formations. The conjectural and past conjectural are no longer generally used productively in standard or Kansai cNJ, but have been replaced by

394

15 NJ granunar

Table 15.3 cNJ adjectival copulaforms Finite

Nonpast Past (Conjectural) (Past conjectural)

i katta (/raroo) (kattaroo)

Non-finite

Infinitive Gerund Conditional Provisional Concessive

ku kute kattara - kutewa kereba kutemo

Negative

ku na-

' tall'

utsukushi' beautifu I'

takai takakatta (takakaroo) (takakattaroo)

utsukushii utsukushikatta (utsukushikaroo) ( utsukushi kattaroo)

Infinitive Gerund Conditional Provisional Concessive

takaku takakute takaka/tara - takakutewa takakereba takakutemo

utsukushiku utsukushikute utsukushikattara - utsukushikutewa utsukushikereba utsukushikutemo

Negative

takaku na-

utsukushiku na-

takaFinite

Nonpast Past (Conjectural) (Past conjectural) Non-finite

attaching the conjectural fonn of the copula to the nonpast or past: takai daroo (yaroo), takakatta daroo (yaroo). in standard cNJ the infinitive is -ku (takaku, utsukushiku) and the onbin-variant, -u, is not used, except in the super-polite formation (cf. 12.3.3), but it is used in the Kansai dialects (takoo, utsukushuu, etc.), where it is also used in the fonns built on the infinitive (gerund takoote, negative takoo na-, sometimes abbreviated, e.g. tako na-). The main inflected copula fonns are as in Table 15.4. Dewa is often abbreviated to ja. As with the adjectives, the past conjectural is very rare today. The finite fonns are built on a stem da which ultimately derives from de aru (see 12.2.2), probably through de aru > *daru > da. Dais almost absent from LMJ,

15.3 Other new NJ grammatical forms

395

Table 15.4 cNJ copula forms Plain

Polite

Nonpast Attributive Past Conjectural (Past conjectural)

da na-no datta daroo (dattaroo)

desu

Infinitive Gerund Conditonal Provisional Concessive

ni -to de dewa nara(ba) demo

Negative

dewa na-

deshita desho

which instead has dya (< dyant < de ant), and is thought to be an eastern Japanese feature which found its way into the common language used in Edo and from there into standard cNJ. It is in that case not an isolated dialect feature of very long standing, for de am, which is the origin of both da and dya, is an LMJ formation. Note, however, that the difference in the pattern of fusion, de ant> Kansai (western) dyant > dya :: cNJ (eastern) *daru > da, is similar to the difference in the OJ statives (3.1.4.7.2, 5.1): *saki-ar- > (western) OJ sakyer- :: EOJ sakar-. Some parts of Kansai use ja, the direct reflex of LMJ dya, but today Kyoto and Osaka mostly use nonpast ya, conjectural yaroo, past yatta, built on a stem ya, which is usually thought to be a further reduction of LMJ dya, but which as mentioned above may instead originate in a reanalysis of the EMJ particle yar:J:J (cf. 12.1.4). The cNJ polite copula is thought to have originated in further reductions of LMJ de S:J:J, copula gerund de + s:J:J, a severely reduced form of the polite existential and auxiliary verb s:J:JY:J:J (i.e., a polite form of de ar-), cf. 12. 7.2.

15.3

Other new modern Japanese grammatical forms

In addition to the changes in verb and adjective inflection described above, a few other grammatical forms are an important part of cNJ and should be mentioned here: The conjunctional particle shi 'and (moreover)' is thought to be from the adjectival conclusive ending -si which disappeared as an inflectional ending with the concl/adn merger (12.2.1). It was used first with nouns and adjectival infinitives towards the end of LMJ and came to be used with verbs in NJ. The evidential extension rasi- 'seems, appears' developed earlier in Kyoto NJ than in Edo; it is not a direct continuation of the OJ extension

396

15 NJ granunar

rasi- (which went out of use in EMJ, cf. 8.5), but is thought to have developed in early NJ from the derivational morpheme -rasi- which appeared in the second half of LMJ and which derives adjectives meaning '-like, typical of, suitable for' (e.g. (from Vocabulario) aforaxij /a[xl-rasii/ 'foolish', aj:J:J 'fool' > NJ aho); derivational -rasi- still coexists with the extension rasi-. Very recently, a similar course of development has given the extension ppoi (iku ppoi 'seems to be about to go') which is more or less synonymous with rasii and which until recently was only used as a derivational morpheme (mizu-ppoi 'watery'). Finally, the 'hearsay' extension soo-da originates in the LMJ evidential auxiliary -(i)soo-na (12.1.5.3) which continues to be used as an auxiliary today.

16

Eastern dialect features of the standard language

Although most features of standard NJ derive straightforwardly from the language reflected in the sources from the end of LMJ, there are a num her of features of standard NJ which reflect influence of eastern Japanese dialects on the Edo version of the common language which came to form the basis for the standard language. It is interesting to consider what Rodrigues says inArte about the language of Kanto (Quanta lkwantool, described as the area stretching from Mikawa (the southern part of present-day Aichi-ken) eastwards), that is to say, about eastern Japanese before the establishment ofEdo as a political centre (Arte, pp. 612-13). He first introduces Kanto Japanese: Generally, the manner of speaking is rough and sharp. Many syllables are swallowed and not pronounced. Further, there are many coarse words particular to this region which can only be Wlderstood among the people from there.

He then goes on to list a num her of features typical of Kanto Japanese. Other than the use of the particle bei (< be-i the necessitive extension + nonpast adjectival copula), these features are described contrastively, summarized in (1). (1) a. b. C.

d. e. f.

/se/ => Allative particle Shape of intentional Adjectival copula infinitive Onbin of -w base verbs Negative auxiliary

Kyoto [Je] e -:J-:JZe-u -u -nu

Kanto [se] sa -a-nze-ku -Q -nai

(a) describes a sound change that took place earlier in Kanto than in Kansai (loss of palatalization before /e/). The allative particle sa is not used in standard NJ. (c)-(e) are onbin forms, where the western forms are more vocalic and the eastern forms more consonantal; the intentional is no longer used, but the distribution of adjectival and verbal forms is like that today, with the Kanto 397

398

16 Eastern dialect features ofthe standard language

forms being used in the standard language (see 16.1). Finally, it is also the Kanto negative auxiliary which is used in standard NJ (see 16.3). The following features of standard Japanese reflect influence from eastern Japanese dialects: In phonology, them ain difference between standard (Tokyo) cNJ and Kyoto cNJ is the prosodic system (see 7.4). In this case it is clear that the cNJ standard (Tokyo) prosodic system does not reflect the Kyoto based common language, but instead an independent eastern Japanese system whose split from the system reflected in EMJ (Kyoto) and in NJ Kyoto predates the written records noting pitch (late EMJ). This is the main instance of an old eastern Japanese feature which became part ofEdo language and subsequently of the variant of the common language used in Tokyo which became the standard language. In grammar, the use of ka as a general sentence final interrogative particle which is now widespread in all of Japan is quite different from the LMJ pattern of distribution of ka and zo exemplified in 12.6.1.2.2 above and is thought to be an eastern dialect feature. Other eastern features of standard cNJ include onbin-forms (16.1), the copula da (16.2), the negative auxiliary -(a)na- (16.3) and the imperative form of vowel base verbs (16.4). In fact, these are the main morphological isoglosses usually listed as separating western and eastern Japanese, cf. Map 16.1. 16.1

Onbin forms

The use of verbal and adjectival onbin forms is a feature that separates eastern and western Japanese, and also standard (Tokyo, eastern) NJ and Kansai (Kyoto, western) NJ. As mentioned above (15.2), the standard language does not use -u, the onbin variant of the adjectival copula infinitive, whereas Kansai (including Kyoto) and Kyushu do. With the verbal onbin stems the situation is more complex. In cNJ, the onbin stem is used in the formation of the past (-(I)ta), gerund ( -(I)te), conditional (-{!)tara), and representative (-(I)tari). The forms in (2a) are mostly shared between western and eastern cNJ. Where LMJ exhibited variation between vocalic and consonantal forms (for -w, -b, -m base verbs), they are now divided as in (2b), which also includes -s bases where eastern cNJ uses a form identical with the infinitive whereas western cNJ retains use of a distinct onbin stem. The very rough basic pattern is that eastern and standard cNJ uses consonantal and western cNJ uses vocalic variants. However, there is today a great deal of variation in the shapes and use of verbal onbin stems within western Japan. Thus for -b and -m base verbs, most of Kansai, including Kyoto, have come to use consonantal, standard language, forms (yonde), whereas far western Honshu and Kyushu use vocalic forms (yode). For -s base verbs, most of Kansai uses the vocalic onbin form (daite) but Kyoto uses the standard non-onbin form (dashite). For -w base verbs

16.1 Onbin forms

399

•

\•

. ~ ), .. . II ~\~· ~~~ '/

::r

Ill

a

~

Ill

-'~

~·\i>

~·,2

.

Ill

0 "'

~.

0

0

•

:J:

~I

0

~.\

i:~-

:J '

0

~\~ Q...•o

~:~ ~·-· ~·:J

- ' •0 0 0

~~ .., .... ....s:: Ill

I -·10. Ill Ill I

\• \

-:1

25

25

50

75 50

100

125 75

150 175

200

100

125

225

Map 16.1 Major east-west isog1osses (from Shibatani 1990: 197)

250 km 150 miles

400

16 Eastern dialect features ofthe standard language

western cNJ generally uses the vocalic (kate) and standard cNJ the consonantal (katte) form. For an overview of the distribution of onbin forms in cNJ, see maps 92-105 (verbal forms) and 137-9 (adjectival forms) in the Grammar atlas ofJapanese dialects (Kokuritsu Kokugo Kenkyujo 1989-2006). In addition, there are slight irregularities for individual verbs; for example, in standard cNJ, the onbin stem of ik- 'go' is iQ- (past itta), not expected *il- (*iita), or tow- 'ask' is toV- (tota), not *toQ- (*totta). (2)

Basic stem a.

mot- 'hold' tor- 'take' kak- 'write' kog- 'row' shin- 'die'

Onbin-stem Western Eastern moQtoQkalkolshiN-

b.

kaw- 'buy' yob- 'call' yam- 'read' das- 'put out'

koVyoVyoVdal-

kaQyoNyoNdashi-

Gerund Western Eastern motte totte kaite koide shinde kate yade yade daite

katte yon de yon de dashite

It should be noted that there is more variability in this respect in the LMJ sources than in either of standard and Kyoto Japanese and that all of the forms used in both standard and Kyoto cNJ are represented in the LMJ sources. However, it should also be noted that Rodrigues, as mentioned above, describes the use of the standard NJ adjectival and verbal onbin as typical ofKanto, and that the Christian sources from the very end ofLMJ use forms that correspond to Kansai NJ. Thus, while we cannot say that the NJ distribution necessarily is of a very long standing, it seems to have been well established from the end ofLMJ.

16.2

Copula da

The paradigm of the standard cNJ plain copula da in Table 15.4 above directly reflects a number of LMJ forms, but also includes new forms in da(-) which were either absent or extremely rare in LMJ (12.2.2). These forms are like the equivalent LMJ forms thought to reflect contraction of dear- such as de aru > de a > da, rather than the usual LMJ outcome dya. This, too, is thought to be an eastern dialect development, and as mentioned above (15.2), this development is similar to the EOJ pattern of phonological fusion in the morphological stative: pre-OJ *saki ar- >(western) OJ sakyer- :: EOJ sakar-, contra.

16.3 The negative auxiliary

16.3

401

The negative auxiliary

In standard cNJ the negative auxiliary is -(a)na- which inflects like an adjective and has the main fonns in (3). The conjectural is not used productively today, but was in use until the Meiji period. (3)

Finite Nonpast Past (Conjectural)

nai nakatta (nakaro)

yomanai yomanakatta (yomanakaro)

tabenai tabenakatta (tabenakaro)

Non-finite Infinitive Gerund Conditional Provisional Concessive

naku nakute nakattara nakereba nakutemo

yomanaku yomanakute yomanakattara yomanakereba yomanakutemo

tabenaku tabenakute tabenakattara tabenakereba tabenakutemo

This formation has in cNJ entirely replaced the LMJ negative auxiliary (12.1.5.1), which is, however, reflected inKyotoNJ and in many other dialects. Rodrigues mentions negative -(a)na- as a feature of Kanto Japanese, giving examples such as aquenai /agenai/ 'doesn't raise', yomanai 'doesn't read' and narauanai /narawanail 'doesn't learn' (Arte, p. 612). This shows that this negative fonnation was in use already at the end ofLMJ and was not a form which emerged in Edo. It is sometimes speculated that the negative -(a)na- is related to the EOJ negative -(a)nap- or -(a)nani (see Chapter 5), but it is more likely that it reflects the adjective na- grammaticalized from a free (grammatical) word to an auxiliary -(a)na-. A possible source for that is a construction such as (4a.ii) which would be the negative counterpart of (4a.i), which has the existential verb ar- used in a light verb type construction in order to topicalize the lexical verb~ (4a.ii) uses the adjective na- as a suppletive negative fonn of the existential verb ar-. The constructions in (4a) are not attested, and the LMJ common language equivalent would be yomi wa se-, using se- 'do' as the light verb. However, EMJ or LMJ eastern dialects are not attested in writing, and it is quite possible that constructions such as (4a) were in use there. In late LMJ na- had become the regular suppletive negated fonn of ar- and in the late LMJ common language sources there are several analytic fonns and constructions with ar- whose negative counterparts use na-, e.g. (4b) analytic adjectival copula fonns (12.2.1), (4c) regular copula fonns (12.2.2), (4d) the respect construction (12. 7.1.2), or ( 4e) the polite existential (12. 7.2). The existence of such constructions may have reinforced the reinterpretation of na- as an auxiliary. The attachment of this auxiliary to the a- stem could in part have been motivated by the conventional attachment of negative fonns to the a- stem

402

16 Eastern dialect features ofthe standard language

serving as an analogical model, but also here the proposed origin in *yomi wa na- fits well and would involve reduction of li wa/ to /a/, reinterpreted as forming the a- stem: *yomi wa na- > *yomya na- > yoma-na-. (4) a.

b. C.

d. e.

(i) *yomi wa ar-

(ii) *yomi wa na- (>*yomya na- > yoma-na-)

taka-(k)u arde(wa) aro-yomi-argoza-ar-

taka-(k)u nade(wa)-nao-yomi-nagoza-na-

Note that the Kansai negative, in addition to the negative nonpast -n (yoman 'doesn't read') which reflects the LMJ negative auxiliary-(a)n-, has developed a new negative, -(a)hen(-) which attaches to the a- stem of consonant base verbs, e.g. yoma-hen 'doesn't read' (yom- 'read'). This new Kansai negative is thought to have developed in a way that is similar to that which we propose as the origin of the Kanto negative, namely through a light verb construction: thus yomahen derives from yomi wa sen, the negative form of the usual light verb construction with se- 'do', through (a) reduction of li wa/ to /a/, reanalysed as forming the a- stem: yomi wa > *yomya > yoma-; and (b), weakening of /s/ to /hi, sen > hen, and reinterpretation of hen as a simple suffix. This innovative negative has in some varieties of Kansai Japanese acquired vowel harmony variants, using /hen/ with consonant and -e base verbs, but lhin/ with -i base verbs: yoma-hen, tabe-hen 'doesn't eat', but oki-hin 'doesn't arise'. Finally, note that both-nand -hen add -katta to form the past tense (yomankatta, yomahenkatta 'if someone didn't read') and -kattara to form the conditional (yomankattara, yomahenkattara 'if someone doesn't read'), both extracted from the paradigm of the adjectival copula.

16.4

The imperative

While both the choice of onbin stem and the replacement of the LMJ negative auxiliary in the standard language were influenced by eastern dialect features, the shape of the imperative is the only feature of standard cNJ which directly reflects an attested dialect feature of long standing. Thus, the shape of the imperative in vowel base verbs in standard cNJ corresponds to the EOJ vowel base imperative (cf. 5.2). The present-day Kyoto forms are usually thought to be the infinitive (usually elongated). (5) akeoki-

OJ ake(yo) oki(yo)

Kyoto cNJ (ake.) (oki.)

EOJ akero okiro

Standard cNJ akero okiro

17

The westernization of Japanese: Loanwords and other borrowings

Particularly since the Meiji period (1867-1912), Japanese has taken in a large number of loanwords from European languages, especially from English. However, Japan's initial contact with European languages is reflected in borrowings from as early as the sixteenth century. The ongoing impact on the Japanese lexicon is of a magnitude equal to the impact Chinese had on the language centuries earlier (see Chapter 9), not simply providing new words for new things and technology, but making available a whole intellectual and philosophical conceptual world which was unknown in Japan before the Meiji period. The intake of loanwords may be divided into three main waves: (a) pre-Meiji, (b) from Meiji to the end ofWWII, and (c) from the end of the war, of which it is the latter two phases which have exerted profound influence on the language.

17.1

Vocabulary layers and hybrid words

Often the existence of three or four different 'vocabulary layers' is posited in descriptions of Japanese: native Japanese words (wago fiJ~, or yamatokotoba), Sino-Japanese words (kango ~~), (recent) loanwords (gairaigo :9'1-*~), and sometime also mimetics (giseigo ;JtEyl'f~, gitaigo -~~). The first three, the main lexical layers, overtly refer to historical origin, but this classification is in addition thought by many scholars to represent some synchronic linguistic reality for speakers of the language, and other defining criteria than etymology are invoked, for example phonological or sociolinguistic ones. However, an etymological classification of words or morphemes is highly relevant for the history of the language, but there is no one-to-one relationship between the etymologically defined layers and layers defined by other criteria. There are certainly phonological features which are exclusive to some recent loanwords (14.6), but they are not shared by all recent loanwords. And there are socio-linguistic values (17.1.1) which are characteristic of recent and not well-integrated loanwords, but this is a short-lived quality which changes if and when the words become better integrated~ more generally this applies to much new vocabulary, regardless of its origin. On the 403

404

17 The westernization of Japanese

whole it seems that the notion of vocabulary layers, other than those defined strictly etymologically, is not useful for a book such as this one. So-called 'hybrid' words, i.e. words composed of elements from more than one etymological layer, abound. We already saw examples of combinations of borrowed SJ vocabulary with preexisting material in 9.2.3.2, and a simple further example is the name of Yanase Takashi's popular cartoon character anpanman, lit. 'bean-jam bun man', which was coined by attaching the male superhero suffix -man borrowed from English to the compound anpan 'beanjam bun', which was coined in the Meiji period from SJ an 'bean-jam' (i§) and pan 'bread' (from Portuguese). Anpanman is thus hybrid in combining three different layers of borrowed material. 17.1.1

Gairaigo

The Japanese word usually translated into English as 'loanword' is gairaigo (71-~!H§), but gairaigo is different from loanword in several respects. First of all, gairaigo is used about recent, transparent borrowings, primarily from European languages, but does not refer to SJ vocabulary, which also is borrowed, or to old naturalized loanwords, such as those mentioned in Chapter 4. The word gairaigo was coined during the period of modernization of the Japanese lexicon which took place in the Meiji period (17.3) and was used first by the linguist Ueda Kazutoshi in 1895 (after returning from his studies in Germany in 1890), and it seems clear that gairaigo corresponds to and is intended to render the Germanfremdwortwhich literally means 'alien, foreign word'. This word, and Japanese gairaigo, captures better than Japanese shakuyogo (1-ti'J¥.1~), German lehnwort, or English loanwonithe various and quite diverse socio-linguistic nuances associated with the use of loanwords from European languages (such as 'learned, trendy, exotic' and others), at least in the initial phase of their use. Words considered gairaigo are usually written in katakana and they are sometimes referred to as katakana-go 'katakana words'.

17.2

Pre-Meiji; from the end of Late Middle Japanese to the middle of the nineteenth century

During the so-called 'Christian century', from the arrival of the missionaries in the 1540s to the ban on Christianity and expulsion of the missionaries in the 1630s, loanwords were mainly taken in from Portuguese. Later on, during the period from the 1640s to the opening of Japan in the 1850s, when contact with Europeans was almost limited to the Dutch or took place through the Dutch settlement on Dejima off Nagasaki, loanwords were taken in from Dutch. As we saw above, the Jesuits freely used Portuguese (and Latin) words

17.2 Pre-Meiji

405

in their publications in Japanese, either spelling them as in the source language in the alphabet texts, or adapting them into kana in the texts in Japanese script (10.2.2), but many of these words never became part of the Japanese language. The following is a short list of loanwords, or foreign words, used in Japanese, listed by Thunberg (1792: 268-9). 1 Certainly, the list comprises words which must have been and remained foreign words, used for convenience in conversation with the few interpreters and others who had contact with the Dutch in Nagasaki, but it is also striking that a fair number of words in the list have survived into and are still used in cNJ (shown in boldface). The list gives an interesting glimpse of words which a visitor at that time such as Thunberg was exposed to. For some of these words, Thunberg cites a source language, here noted as 'h(olland)' for Dutch' and 'p(ortugal)' for Portuguese, and it is noteworthy that of the words surviving into cNJ, more than half are from Portuguese. Several of the words listed involve both borrowed and Japanese elements (the latter here written in plain type). (1)

tabaco (p) 'tobacco' (cNJ tabako, mainly used about cigarettes), bir (h) 'beer' (cNJ biiru), Portugal abra 'olive oil' (abura 'oil'), unicorn 'unicorn's horn', boter (h) 'butter', boter no kas 'cheese' (no genitive), karta (p) 'playing cards' (cNJ kamta), kapitein (h) 'captain', lancetta 'lancet', krokodil 'crocodile', pang (p) 'bread' (cNJ pan), pokk (p) 'venereal disease', savon (p) 'soap' (cNJ shabon, now not used much, but generally replaced by SJ sekken .:5~), fige savon 'shaving soap' (fige > cNJ hige 'beard'), fiasco (p) 'flask', banco (p) 'bench', diamant 'diamond',faka 'knife', saffran 'saphron', biduro (p) 'glass', baso 'dish', kopp (h) 'cup' (cNJ koppu), ducaton 'Dutch coin', theriak 'theriac', gans (h) 'goose', vein (h) 'wine' (cNJ wain), kakami biduro 'looking glass' (kagami 'mirror'), rassia, orassia 'cloth' (o- Honorific prefix; cNJ rasha; Thunberg does not note it, but this word is from Portuguese raxa), aderlaten (h) 'blood-letting', skerbekken (h) 'basin for shaving',fork (h) 'fork' (cNJ hooku,fooku), kananor isi 'hematite' (Cannanore (Kannur) in India+ isi 'stone'), tabaco ire 'tobacco pouch' (ire 'receptacle'), tinta 'Spanish wine'.

Other Portuguese loanwords which have survived into cNJ include bateren 'priest, padre'(< padre), kasutera 'sponge cake'(< (pao de) Castella 'bread 1

The list is introduced by: 'Sunt pneterea res, in Japoniam ab Exteris allatre, qure nominibus vel Portugallicis vel Hollandicis insigniri so lent, ut:' ('there are various things, brought to Japan from outside, which are usually designated by Portuguese or Dutch names, such as:').

406

17 The westernization of Japanese

from Castilla'), and tenpura 'tempura~ deep fried fish and vegetables in batter' (
17.3

From the beginning of Meiji to the end of WWII (1945)

Although the Dutch officially until the middle of the nineteenth century were the only Europeans allowed in Japan, there was some contact with other Europeans and their languages from the beginning of the nineteenth century, but it was not until the enforced opening of Japan in the 1850s that loanwords from other European languages than Portuguese and Dutch were taken in to any noteworthy extent. From the middle of the nineteenth century Japan was inundated with technology, science, ideas and material culture from Europe 2

There are also words for 'pumpkin' which have different origin, such as place names presumably reflecting putative origins (nankin after Nanjing in southern China, or chosen 'Korea') or descriptions (t6-nasu 'Chinese eggplant').

17.3 From the beginning of Meiji to the end ofWWII

0

kabocha b6bura - bobura

bobuna bun a 6 b6bora - bobora ~ bon bora bofura - bofura dofura I> b6ran 0

®

~

•• 0

tobora

L nankin

T ~

v

nankwa(n) nanku t6nasu chinkwa nan ban

T

bontan bonkan 1r t6chiburu yugo

y

z.

... l

if

c

satsuma chosen

~
kinto tuss6 toppy6

~

1(/

J Map 17.1 Distribution of words for 'pumpkin' (from Sato 2002: 169)

407

408

17 The westernization of Japanese Table 17.1 Loanwords in use during the Taisho period (from ShibaJani 1990: 149, table 7.5) Origin

Number of words

%

English Dutch Portuguese French German Spanish Others

84 45 23 6 2

51.9 27.8 14.2 3.7 1.2 0.6 0.6

and the US, which was eagerly adopted in the course of the effort to modernize and catch up with the 'western' countries. This resulted in a great amount of borrowed vocabulary, which was taken in either as direct loans (17.3.1), or as loan translations into SJ (17.3.2).

17. 3.1

Loanwords from European languages

English dominated as a donor language of loanwords from the very beginning of the period of modernization, but it did not attain the virtual monopoly it has today until after the end ofWWII. It is often observed that loanwords from the minor donor languages exhibit some degree of semantic specialization, correlating with areas of material and cultural contribution: e.g. German (karute 'medical patient record'< karte, (aru)baito 'part-time student job' < arbeit), French (bifuteki 'steak' < bifteck, konsome < consomme), Italian (opera, sonata). Table 17.1 shows the origins of 162loanwords from different European languages which were in common use around the middle of the Taish6 period (1912-26). It is noteworthy that words borrowed from Dutch and Portuguese still at this time were quite prominent.

17.3.2

Loan translations; Sino-Japanese coinages

In addition to direct loans, a strategy was adopted of coining SJ words for the new notions, institutions and things to be named, that is, loan translation into (Sino-)Japanese. This coinage often consisted of finding some word, or kanji combination, in one of the Chinese Classics, which could be drafted in to write the new word, which was vocalized by using the SJ reading, or alternatively of reviving or adapting SJ words from earlier or specialized usage, but freer coinage from Sino-Japanese was also frequently used.

17.3 From the beginning of Meiji to the end of WWII

409

Some of the new SJ words originate in kanji writings intended as a logographic representation of a direct loanword, accompanied by reading glosses in katakana, subsequently reinvented as a SJ word. An example which illustrates some of the processes involved is cNJ shokudo 1t~ which is now the general word for most types of '(large, public) dining room, dining hall'. In the 1870s English dining room was rendered or borrowed as daininguriimu, written as ~~jlt (·)l-1::::. :/ -!1 ;J...-~ A), i.e. logographically in kanji with the pronunciation shown in katakana. Earlier ~jlt zikidoo < zikidafi (which is go-on), was a specialized word for dining rooms in Buddhist temples, but following on from the writing of daininguriimu as shown above, ~jlt was drafted in as a SJ translation equivalent and was, accompanying its new use, given a different pronunciation (reading) using kan-on for ~, but keeping go-on for jlt (whose kan-on is to). Incidentally, daininguriimu is still 1/.(:::.<'~/1#--A

occasionally written ~ ~ . SJ words deliberately coined, revived or adapted in this way during the Meiji period make up the great majority of SJ words in use today - and the overwhelming majority of academic, political and intellectual vocabulary used in Japanese today -and some of these words were later adopted in China and Korea, with Chinese or Sino-Korean vocalization. Many of these new words are attributable to individuals who also otherwise played prominent roles in the process of modernization in the nineteenth and early twentieth century, such as Nishi Amane, Fukuzawa Yukichi and Inoue Tetsujiro, whose translation of William Fleming's The vocabulary of philosophy, mental, moral and metaphysical (1857), Tetsugaku jii ('fg~*;i;, 1881) provides more than 2,500 SJ words relevant to philosophy. Of the many new SJ words coined in this way in the late nineteenth century, a few examples are given in (2): (2)

shakai t±~ 'society', minshu(shugi) .B::::£.(:3::.~) 'democracy', jiyu § ~ 'liberty', shinri(gaku) 'L'l!!!.(~) 'psyche (psychology)', tetsugaku 'fg~ 'philosophy', kaisha ~t± 'firm, company', kaikeigaku ~*~ 'accounting', rodo jj-IJ 'work' (which includes -1! which was made in Japan, see 9.1.5), yubin l!Qi19! 'post',jidosha§lil.$: 'automobile', tetsudo ~~'railway', denwa ~~'telephone', enzetsu ~~'speech', toron W1~ 'discussion', bungaku X~ 'literature', shosetsu 1j'~ 'novel', kokka 00* 'nation', kokumin OOB:: 'people'.

As the short list suggests, new words for abstract concepts and institutions were usually coined in this way, rather than by direct loans. It is highly unlikely that the lexical, terminological modernization of the Meiji period would have been as successful as it was, if it had had to rely on direct loans, rather than

410

17 The westernization of Japanese

SJ coinages, which are short and in some cases provide educated readers with semantic clues. In addition, SJ coinages have the important function of giving the impression that these words were part of and belonged in an intellectual tradition. 17. 3. 3

Influence from European languages in grammar and usage

European languages, and especially translation practices from European languages not unlike the kanbun-kundoku practices described in 9.1 -i.e. 'reading' a Dutch or English text 'in Japanese' - have exerted some influence on Japanese, though not to the extent Chinese did earlier. For example, the use of the 3rd person pronouns kare 'he' and kanojo 'she'(< ka-no-jo 'DISTALGEN-woman'), which became widely used from the late nineteenth century in literary writing and eventually passed over into general language, originates in translation-inspired imitations of European literature; the imitation went so far as to use kanojo to refer to ships. In general, under this influence from Dutch and English, pronouns came to be used far more, especially in writing from the early twentieth century onwards, than they had been before the Meiji period. Another example which is often mentioned is changes in the use of the passive. Because of the relatively free word order, a semantic patient can in Japanese be topicalized or fronted without the need for passivization, and the passive is traditionally said not to have been used much solely for this purpose, but usually to express some kind of affectedness on a human, or animate, patient, by a sentient or personified agent. However, since the Meiji period, more direct uses of the passive have become current, with both inanimate agents and patients, and this is usually ascribed to influence from or imitation of translation from Dutch and English languages. Kinsui (1997) has shown that marking of the agent in direct passives by ni-yotte originates in translation practices from Dutch, in which door 'through', which can be used both about path, means and way, and about agents in passive constructions, was conventionally translated by ni-yorite or ni-yotte, which originally meant 'because', also when used about passive agents. This usage was eventually generalized and made its way into general language. It is interesting that the title of Yamada Yoshio's ground-breaking study of the influence of kanbun-kundoku on Japanese (Kanbun no kundoku ni yorite tsutaeraretaru goho from 1935), uses the direct passive with an inanimate agent and ni-yorite to mark the agent, both of which originate in similar influences from similar practices centuries after kanbun-kundoku made its mark on Japanese. Finally, the obligatory marking of all core arguments (subjects and objects) by case particles in written Japanese today is a written language feature which was introduced in the establishment of the new normative standard written

411

17.4 Post-WWII Table 17.2 Loanwords used in magazines (from ShibaJani 1990: 148, table 7. 4) Origin

Number of words

%

English French German Italian Dutch Russian Chinese Portuguese Spanish Others

2,395 166 99 44 40 25 22 21 21 131

80.8

5.6 3.3 1.5 1.3 0.8

0.7 0.7 0.7 5.3

language. It was not a feature of written Japanese before the genbun 'itchi reforms, nor was it ever, or is today, a feature of spoken Japanese, where omission of case particles (case drop) has always been frequent. 3 Its introduction into standard written Japanese is usually ascribed to a desire to have a normative, regular grammar for written language, as the European languages did, but it should also be noted that kuntengo (9.1.6), which is one formal genre of written Japanese, generally did not have case drop.

17.4

Post-WWII

The period after the end of WWII has seen an intake of loanwords from English on a much larger scale than in the period before the war and English now has a near monopoly as a donor of loanwords to Japanese, as shown in Table 17.2 which shows a clear growth in the proportion of loanwords from English in the post-war period compared to before the war, cf. Table 17.1. This has accompanied improvements in English education and widespread command of English (i.e., more (limited) bilingualism) and far greater contact with and access to English, in particular American English. It is today difficult to distinguish between integrated loanwords from English and occasional use of phonologically more or less adapted English words (which may later be more widely adopted to become well-integrated loanwords). Conventions exist for spelling English words in katakana, and it is therefore easy to insert English words in written Japanese for stylistic effect. 3

That is the reason a trick question likepantsukuttakoto aru? (courtesy of Miss Saito Sachiko) works. It can either be parsed as (i) pan (o) tsukutta koto aru? 'Have you ever made bread?' or as (ii) pantsu (o) kutta koto aru? 'Have you ever eaten pants?'

412

17 The westernization of Japanese

Needless to say, different speakers have different repertoires of integrated loanwords. In this respect the socio-linguistic situation in Japan is much the same as in most other developed countries, with quite different levels of command of English between different segments of the population and widespread use of English words by some speakers, especially educated and young urban people. Buzzwords often derive from English, for example sekuhara 'sexual harassment', abbreviated from sekushuaru harasumento, the phonologically adapted rendition of the English word; and inventive metaphorical usages can be surprising and amusing, as in biikodo 'comb-over' which is from bar code. English words and morphemes are in Japan often used creatively in combinations that are not usual in first language English speech communities, giving rise to 'loanwords' which are not used in English, for example oeru 'female office worker'< OL (short for office + lady), batontatchi 'baton pass' or as a verb batontatchi-se- 'pass the baton to one's successor'< baton + touch, sukinshippu 'physical contact (especially between baby and mother)' <skin+ -ship, or makaroni-wesutan 'spaghetti western'. Camcorder (Japanese kamukodii.) is an example of an English word coined in Japan and now used worldwide. These examples are all from the post-war period, but many examples are found also before the war, for example orudomisu 'old maid'< old+ miss, or afureko 'dubbing'< after+ recording. REFERENCES Westernization, general: Miller 1967: 235fT., Shibatani 1990: 142fT. Loanwords: Kadokawa gairaigojiten (Arawaka 1977). Grammatical borrowing: Kinsui 1997.

APPENDIX

Summary of the main regular phonemic changes between OJ and conservative N J

This table exemplifies the main regular phonemic changes which have taken place between OJ and conservative NJ (see Chapter 14). It does not include the onbin sound changes (see 7.1.4), which were not regular sound changes, nor phonetic changes which have not resulted in phonemic change. The first row includes cross-references to the main text. In the first column is given a tag gloss, not reflecting changes in word meaning which may have occurred over time. Most words are given in their OJ shape in the second column, but SJ vocabulary or kanji readings are entered at around 950 in the shape they would have had then, although not all are attested from that time.

413

414

Appendix

Gloss

OJ

Loss of ko-otsu distinctions (7.3.2.1)

Merger of/.ye/ and/.e/ (7.3.2.2)

Merger of medial/pi and/w/ (7.3.1)

Loss of syllable initial /w/ before /o, i, e/ (7.3.2.3; 11.2)

Merger ofll, 0/ and /I, U/ (11.1.2)

< 800

< 950

c. 950

950-1000

c. 1000 c. 1100

c. 1300

Cye > Ce Cwi>Ci

Cwo> Co

.ye> .e

/-p-/ > 1-w-1 _li, e, a, o/ /-p-/ >0/ lui

.w> 0

initial position #.w>0 _ji, e/

early LMJ !iJ > /JJ lUI> /U/

'front'

mapye

mape

_jol

medial position -.w>0 _li, e/ mae

mawe

'today'

kyepu

kepu

'love'

kwopwi

kwopi

keu

'shell'

kapi

'indigo'

awi

ai

'voice'

kowe

koe

kopi

'river'

kapa

'branch'

yeda

eda nue

kowi

koi

kawi

kai

kawa

'thrush'

nuye

'hill'

woka

oka

'blue'

awo

ao

'face'

kapo

'well'

wi

i

'smile'

wemi

erni

'field'

para

'plain'

pira

'boat'

pune

'morning

asapi

kawo

kao

sun' 'the road to Ki'

kidi

'classic' (IIi!£)

kei

'sulra' (IIi!£)

kyail

'attachment' (t'l,)

sipu

'need' (!ll!)

eu

'public' (-0)

koil

'fruit; cookie' (:lfi:-1-)

kwasi

'first year in a dynastic period'

(5G1f.)

gwannen

kei kyau siu

kou

Appendix

415

Fricativization of/p/ (11.3)

Monophthongization of NUl (11.5)

Merger of /dl and /z/ before /i, u/ (14.1)

Merger of !CXJ! and fool (14.2)

Delabialization of If/ (14.3)

Loss of /w/ after /k, g/ (14.4)

early? LMJ

mid? LMJ /iU/ > /yuu/ /eU/ > lyoo/ /aU/> /uu/ /aU/>h':>/

17th century IMI> fool

?1700

/p/ >Iff

17th century d > zl _{i, u}

late 19th century /kwa, gwa/> /ka, ga/

If/> !hi

Conservative NJ

mae kyo a

kyo koi kai ai koe kawa eda nue aka ao kao i

emi far a

hara

hara

lira

hira

hira

fune

hune

fune

asafi

asahi

asahi

I

kiji

kizi

kei ky<><>

kyoo

kyo

syuu

syii

yoo

yo

koo

ko kasi

kashi

gannen

gannen

References

Arakawa, Sobe (ed.). 1977. Kadokawa garaigo jiten, 2nd edn. Tokyo: Kadokawa Shoten. Bentley, John R. 2000. A new look at Paekche and Korean data from Nihon shoki. Language Research 36/2, 417-43. 2001 a. A descriptive grammar ofEarly Old Japanese prose. Lei den: Brill. 200lb. The origin ofman'y0gana. Bulletin ofthe School ofOriental andAfrican Studies 64/1, 59-73. Collado, Diego. 1632a. Ars grammaticae iaponicae linguae. Rome. 1632b. Dictionarium sive thesauri lingure Iaponicre compendium. Rome. Comrie, Bernard. 1976. Aspect: An introduction to the study of verbal aspect and related problems. Cambridge: Cambridge University Press. de Boer, Elisabeth M. forthcoming. The historical development of Japanese tone; volume 1: The tone systems of the modem dialects; volume II: The introduction and adaptation of the Middle Chinese tones in Japan (Veroffentlichungen des Ostasien-Instituts der Ruhr-Universitiit Bochum). Wiesbaden: Harrassowitz. DeFrancis, John. 1984. The Chinese language: Fact and fantasy. Honolulu: University of Hawai 'i Press. Doi, Tadao (trans.). 1955. Nihon daibunten [A Japanese translation of Rodrigues 1604-8]. Tokyo: Sanseido. Doi, Tadao, Morita Takeshi and ChOnan Minoru (eds. and trans.). 1980. (Hoyaku) Nippojisho [A Japanese translation with introduction and postscript of Rodrigues 1603-4]. Tokyo: lwanarni Shoten. Durst-Andersen, Per. 1996. Towards a multifunctional grammar: 'Language, reality and mind' in a grammatical description. Hermes, Journal of Linguistics 17, 75-102. Elliott, Ralph W. V. 1959. Runes: An introduction. Manchester: Manchester University Press. Endo, Orie. 2006. A cultural history ofJapanese women slanguage. Ann Arbor: Center for Japanese Studies, the University of Michigan. Engberg, Jonas. 1990. Observationes in linguam japonicam: Observations on the Japanese language [Introduction to and translation ofThunberg 1792]. Stockholm: Center for Pacific Asia Studies, University of Stockholm. Farris, William Wayne. 1998. Sacred texts and buried treasures. Honolulu: University of Hawai 'i Press. Fleming, William. 1857. The vocabulary ofphilosophy, mental, moral and metaphysical. Glasgow. 416

References

417

Frellesvig, Bjarke. 1995. A case study in diachronic phonology: the Japanese onbin sound changes. Aarhus: Aarhus University Press. 1998.Accent, word tone andmetatonyin Central Japanese. On 'inKenkyU/Phonological Studies (published by the Phonological Society of Japan) 1, 195-202. 200 1. A common Korean and Japanese copula. Journal ofEast Asian Linguistics 10, 1-35. 2007. On the verb morphology of Old Japanese. In Current issues in the history and Structure of Japanese, ed. Bjarke Frellesvig, Masayoshi Shibatani and John Charles Smith, 219-252. Tokyo: Kurosio Publishers. 2008. On reconstruction of proto-Japanese and pre-Old Japanese verb inflection. In Proto-Japanese: Issues and prospects, ed. Bjarke Frellesvig and John Whitman, 175-92. Amsterdam: John Benjamins. Forthcoming. Chronological layering in proto-Japanese and pre-Old Japanese verb inflection. In Studies in Japanese and Korean Linguistics, ed. Bjarke Frellesvig, Jieun Kiaer and Janich Wrona. London: Lincom. Frellesvig, Bjarke, Jieun Kiaer and Janick Wrona (eds.). Forthcoming. Studies in Japanese and Korean linguistics. London: Lincom. Frellesvig, Bjarke, Masayoshi Shibatani and John Charles Smith (eds). 2007. Current issues in the history and structure ofJapanese. Tokyo: Kurosio Publishers. Frellesvig, Bjarke and John Whitman (eds.). 2008a. Proto-Japanese: Issues and prospects. Amsterdam: John Benjarnins. 2008b. Evidence for seven vowels in proto-Japanese. In Proto-Japanese: Issues and prospects, ed. Bjarke Frellesvig and John Whitman, 15-41. Amsterdam: John Benjamins. Forthcoming. Prenominal complementizers and the derivation of complex NPs in Japanese and Korean. Japanese/Korean Linguistics 18. Fukuda, Yoshisuke. 1965. Nara jidai Azuma hagen no kenkyii. Tokyo: Kazama Shoin. Hamada, Atsushi. 1970. Chosen-shiryo ni yorn Nihongo kenkyU. Tokyo: Iwanami Shoten. Harbsmeier, Christoph. 1981. Aspects of classical Chinese syntax. London: Curzon Press. Hashimoto, Shiro. 1966. Kodaigo no shiji-taikei: Jodai o chfisin ni. Kokugo Kokubun 3516, 329-41. 1982. Shijigo no shiteki tenkai. In Koza nihongogaku, vol. II: Bunposhi, ed. Kenji Morioka eta/., 217-40. Tokyo: Meiji Shoin. Hattori, Shiro. 1978-9. Nihonsogo ni tsuite 1-22. Gekkan Gengo 7/1-7/3, 7/6-8/12. 1979. Ono Susumu 'Doshi katsuyokei no kigen' ni tsuite. In Nihon no gengogaku, vol. IV (Bunpo II), ed. Hattori Shiro, Kawamoto Shigeo and Shibata Takeshi, 658-63. Tokyo: Taishiikan Shoten. Hayata, Teruhiro. 1998. Jodai nihongo no onsetsu kow to o-retsu ko-otsu no betsu (The syllable and the distinction between o 1 and o2 in Old Japanese). Onsei KenkyU!Journal of the Phonetic Society ofJapan 2, 25-33. 2000. The liquid and stem-final vowel alternations of verbs in ancient Japanese. GengoKenkyii 118,5-27. Hida, Yoshifumi, Endo Yoshihide, Kato Masanobu, Sato Takeyoshi, Hachiya Kiyoto, Maeda Torniyoshi (eds.). 2007. Nihongogaku kenkyi1jiten (!'he research encyclopedia ofJapanese linguistics). Tokyo: Meiji Shoin.

418

References

Horrocks, Geoffrey. 1997. Greek: a history of the language and its speakers. London: Longman. Imaizumi, Tadayoshi. 1971. Nippojisho no kenkyU. Tokyo: Ofiisha. Inoue, Tetsujiro. 1881. Tetsugakujii. Tokyo: Tokyo Daigaku Sangakubu. Iwai, Yoshio. 1970. Nihon gohO-shi, vol. I: Nara, Heian-jidai-hen. Tokyo: Kasama Shoin. 1971. Nihon goho-shi, vol. II: Kamakura-jidai-hen. Tokyo: Kasama Shoin. 1973. Nihon goho-shi, vol. III: Muromachi-jidai-hen. Tokyo: Kasama Shoin. 197 4. Nihon goho-shi, vol. IV: Edo-jidai-hen. Tokyo: Kasama Shoin. Iwanami (ed.). 1957-69. Nihon kotenhungaku taikei (IOOvols. + 2 indexvols.). Tokyo: Iwanami Shoten. Kageyama, Taro. 1999. Word formation. In The handbook ofJapanese linguistics, ed. Tsujimura Natsuko, 297-325. Oxford: Blackwell Publishers. Kamei, Takashi. 1954. On the authenticity of on-readings in Sino-Japanese. The Annals of the Hitotsuhashi Academy 5, 97-105. Kanno, Hiroomi. 1978. ChOsengo kara no shakuyogo. Gekkan Gengo 7/2, 36-42. Kariya, Boshi (compiler), Yamada Yoshio, and Katori Hozuma (eds.). 1912. Kokyo ihun. Tokyo: Hobunkan. Kasuga, Masaji. 1985. Saidaiji-hon Konkomyo saishoo kyo no koten no kokugogakuteki kenkyU. Tokyo: Benseisha. Kawabata, Yoshiaki. 1976. Yogen. In Iwanami koza nihongo, ed. Ohno Susumu and Shibata Takeshi, vol. VI (Bunpo I), 169-217. Tokyo: Iwanami Shoten. Kindaichi, Haruhiko, Hayashi Oki, and Shibata Takeshi (eds.). 1988. Nihongo hyakka daijiten {.4n encyclopedia of the Japanese language). Tokyo: Taishukan Shoten. Kindaichi, Haruhiko and Wada Minoru. 1980. Kokugo akusento ruibetsu goi-hyo. In Kokugogakkai 1980, 7-10. King, J. Ross and John Whitman (eds.). Forthcoming. Special issue of Sino-Platonic Papers. Kinsui, Satoshi. 1995. Nihongoshi kara mitajoshi. Gengo KenkyU 24/11, 78-84. 1997. The influence of translation on the historical development of the Japanese passive construction. Journal ofPragmatics 28/6, 759-79. 2006. Nihongo sonzai hyogen no rekishi. Hitsuji Shobo. 2007. The interaction of argument and non-atgument in the diachronic syntax of Japanese. In Current issues in the history and structure of Japanese, ed. Bjarke Frellesvig, Masayoshi Shibatani and John Charles Smith, 253-61. Tokyo: Kurosio Publishers. Kitagawa, Kazuhide. 1982. Shoku nihongi senmyo: Kohon, sosakuin. Tokyo: Yoshikawa Kobunkan. Kiyose, Gisaburo. 1985. Heian-chO ha-gyo shiin P-onron (Evidence that modern Japanese h- wasp- in the Heian period). Onsei no kenkyU 21, 73-87. Kobayashi, Akemi. 1981. Ennin no kijutsu suru sansukuritto onsetsu 'ca' no onka: kyii-seiki nihongo suitei no kokoromi. Osaka Gaikokugo Daigaku Gakuho 52, 63-80. Kokugogakkai (eds.). 1955. Kokugogakujiten. Tokyo: Tokyodo Shuppan. 1980. Kokugogaku daijiten. Tokyo: Tokyodo Shuppan. Kokuritsu Kokugo Kenkyiijo (ed.). 1989-2006. Hogen hunpo zenkoku chizu (Grammar atlas of Japanese dialects), 6 vols. Tokyo: OkurashO Insatsukyoku.

References

419

Komatsu, Hideo. 1956. Nihon jion ni okeru shinnai-nisshO-inbi no sokuonka to zetsunainissho e no goryil-katei. Kokugogaku 25. 1981. Nihongo no on'in (Nihongo no sekai 7). Tokyo: Chao Koronsha. Kono, Rokuro. 1969. The Chinese writing and its influences on the scripts of the neighbouring peoples: With special reference to Korea and Japan. Memoirs ofthe Research Department of the Toyo Bunko 27, 83-140. Kotani, Hiroyasu. 1986. Mokkan to senmyo no kokugogakuteki kenkyU. Osaka: Izumi Shoin. Kubozono, Haruo (ed.). 2008. Asymmetries in phonology: An East-Asian perspective. Tokyo: Kurosio Publishers. Kuroda, S.-Y. 2007. On the syntax of Old Japanese. In Current issues in the history and structure of Japanese, ed. Bjarke Frellesvig, Masayoshi Shibatani and John Charles Smith, 263-317. Tokyo: Kurosio Publishers. Lange, Roland A. 1973. The phonology of eighth-century Japanese. Tokyo: Sophia University. Levy, Ian Hideo. 1981. Man 'yoshU: A translation of Japans premier anthology of classical poetry. Tokyo: Tokyo University Press. Lewin, Bruno. 1976. Der koreanischeAnteil am WerdenJapans. Opladen: Westdeutscher Verlag. Mabuchi, Kazuo. 1971. Kokugo on 'inron. Tokyo: Kasama Shoin. 1993: Gojiionzu no hanashi. Tokyo: Taishiikan. Martin, Samuel E. 1968. On the forms of Japanese adjectives. Glossa 2, 46--69. 1975. A reference grammar of Japanese. New Haven: Yale University Press. 1987. The Japanese language through time. New Haven: Yale University Press. 1990. Morphological clues to the relationships between Japanese and Korean. In Linguistic change and reconstruction methodology, ed. Phili Baldi, 483-509. Berlin and New York: Mouton de Gruyter. 1991. Recent research on the relationships of Japanese and Korean. In Sprung from some common source: Investigations into the prehistory oflanguages, ed. Sydney M. Lamb and E. Douglas Mitchell, 269-92. Stanford: Stanford University Press. Matsumori, Akiko. 2008. On the reconstruction of the proto-accentual system of Japanese. InProto-Japanese: Issues and prospects, ed. Bjarke Frellesvig and John Whitman, 103-24. Amsterdam: John Benjamins. Matsumoto, Katsurni. 1984. Gengoshi no saiken to gengofuhen. Gengo Kenkyu 86, 5-32. Matsumura, Akira (ed.). 1971. Nihon hunpo daijiten. Tokyo: Meiji Shoin. 1998. {Zoho) Edogo Tokyogo no kenkyU. Tokyo: Tokyodo Shuppan. Miller, Roy A. 1967. The Japanese language. Chicago: University of Chicago Press. 1971. Japanese and the other Altaic languages. Chicago: University of Chicago Press. 1989. Old Japanese i. In Bruno Lewin zu Ehren: Festschrift aus An/ass seines 65. Gehurtstages, ed. Irmela Hijiya-Kirschnereit and Jiirgen Stalph, vol. I, 251-91. Bochum: Brockmeyer. Miyake, Marc. 1997. Pre-Sino-Korean and pre-Sino-Japanese: Reexamining an old problem from a modem perspective. Japanese/Korean Linguistics 6, 179-211. 2003a. Old Japanese: a phonetic reconstruction. London: RoutledgeCurzon. 2003b. Philological evidence for *e and *o in Pre-Old Japanese. Diachronica 20/1, 83-137.

420

References

Mizushima, Yoshiharu. 1984. Man 'yoshU azuma-uta honbun kenkyU narabi ni si5sakuin. Tokyo: Kasama Shoin. 2003. Man 'yoshii sakimori uta zenchUshaku. Tokyo: Kasama Shoin. Morita, Takeshi. 1993. Nippojisho teiyi5. Osaka: Seibundo Shuppan. Murayama, Shichiro. 1957. Vergleichende Betrachtung der Kasus-Suffixe im Altjapanischen. Studia Altaica: Festschrift fur Nikolaus Poppe, 126-31. Wiesbaden: Otto Harrassowitz. Murayama, Shichiro and Roy A Miller. 1979. The Inariyama tumulus sword inscription. Journal ofJapanese Studies 5, 405-38. Nakada, Norio. 1969. Ti5daiji fujumonki5 no kokugogakuteki kenkyU. Tokyo: Kazama Shobo. Nakada, Norio and Hayashi Chikafumi. 2000. Nihon no kanji. Tokyo: Chuokoron Shinsha. Nelson, Andrew Nathaniel. 1974. The modern reader's Japanese-English character dictionary, second revised edition. Rutland, VT: Tuttle. Nishizaki, Toru. 1992. 'Hokke mongu 'koten no kokugogaku-teki kenkyii: Honbun-hen. Tokyo: Ofu.sha. (ed.). 1995. Nihon kojisho o manabu hito no tame. Kyoto: Sekai Shis0sha. Nomura, Takeshi. 1989. Jodaigo no tu to nu ni tsuite. Kokugogaku 158, 1-14. Numoto, Katsuaki. 1986. Nihon kanjion no rekishi. Tokyo: Tokyodo Shuppan. Ogura, Hajime. 1998. Sa-gyo shiin no rekishi. Kokugogaku 195, 42-55. Ohno, Susumu. 1953. Nihongo no doshi no katsuyokei no kigen ni tsuite. Kokugo to Kokubungaku 350, 47-56. 1977. Kanazukai no rekishi. In Iwanami ki5za nihongo, ed. Ohno Susumu and Shibata Takeshi, vol. VIII (Moji), 301-39. Tokyo: Iwanami Shoten. 1978a. Doshi katsuyokei no kigen. In Nihongo bunpi5 o kangaeru, Ohno Susumu, 1979-222. Tokyo: Iwanami Shoten. (Reprinted 1979 in Hattori Shiro, Ono Susumu 'Doshi katsuyokei no kigen' nit tsuite. In Nihon no gengogaku, vol. IV (Bunpi5 II), ed. Hattori Shiro, Kawamoto Shigeo and Shibata Takeshi, 658-63. Tokyo: Taishiikan Shoten.) 1978b. Nihongo no bunpi5 o kangaeru. Tokyo: Iwanami Shoten. 1980. Nihongo no seiritsu (Nihongo no sekai 1). Tokyo: Chfio Koronsha. 1993. Kakarimusubi no kenkyii. Iwanami Shoten. Ohno, Susumu et al. (eds.). 1990. Iwanami kogo jiten, 2nd, enlarged edn. Tokyo: Iwanami Shoten. Omodaka, Hisataka et al. (eds.). 1967. Jidai-betsu kokugo daijiten: Ji5daihen. Tokyo: Sanseido. Otsuka, Mitsunobu and Kita Takashi. 1999. Esopo no haburasu: Honbun to si5sakuin. Osaka: Seibundo Shuppan. Philippi, Donald L. 1968. Kojiki: Translated with an introduction and notes. Tokyo: Tokyo University Press. 1990. Norito: A translation of the ancient Japanese ritual prayers (with a new preface by Joseph M. Kitagawa). Princeton : Princeton University Press. Pulleyblank, Edwin G. 1984. Middle Chinese: A study in historical phonology. Vancouver: UBC Press. 1991. Lexicon of reconstructed pronunciation in Early Middle Chinese, Late Middle Chinese, and Early Mandarin. Vancouver: UBC Press.

References

421

Quim, Charles Joseph. 1987. A fi.mctional granunar of predication in Classical Japanese. Doctoral dissertation, University of Michigan. Ramsey, S. Robert. 1979a. Language change in Japan and the odyssey of a teisetsu. Journal ofJapanese Studies 8/1, 97-131. 1979b. The Old Kyoto dialect and the historical development of Japanese accent. Harvard Journal ofAsiatic Studies 39/1, 157-75. 1991. The polysemy of the term kokugo. In Schriftfestschrift: Essays on writing in honor ofJohn DeFrancis on his eightieth birthday (Sino-Platonic Papers 27), ed. Victor H. Mair, 37-47. Philadelphia: University of Pemsylvania, Department of Oriental Studies. Rodrigues, Ioao. 1603-4. Vocabulario da lingoa de Iapam. Nagasaki. 1604-8. Arte da lingoa de Iapam. Nagasaki. 1620. Arte breve da lingoa Iapoa. Macao. Russell, Kerri L. 2006. A reconstruction and morphophonemic analysis of protoJaponic verbal morphology. Doctoral dissertation, University of Hawai'i at Manoa. Forthcoming. A grammar of Eastern Old Japanese. Folkestone: Global Oriental. Saeki, Umetomo. 1959. Narajidai no kokugo. Tokyo: Sanseido. Sandness, Karen. 1986. The pronunciation of the s-initial syllables in the Nara period: a critical look atArisaka's and Mabuchi's theories. Papers in E astAsianLanguages 3, 1-16. 1999. The evolution of the Japanese past and perfective suffixes (Michigan Monograph Series in Japanese Studies, 26). Am Arbor: Center for Japanese Studies, University of Michigan. Sato, Kiyoji (ed.). 1977. Kokugogaku kenkyiljiten. Tokyo: Meiji Shoin. (ed.). 1982. Kodai no goi. Tokyo: Meiji Shoin. Sato, Ryoichi (ed.). 2002. 0-kuni kotoba o shiru: Hogen no chizucho. Tokyo: ShOgakukan. Sato, Takeyoshi. 1982. Kodai goi no gaisetsu. InK. Sato 1982, 45-77. Seeley, Christopher. 1991. A history of writing in Japan. Leiden: E. J. Brill. Shibata, Takesi. 1998/1965. Theriseandfallofdialects. InTakesiSibata: Sociolinguistics in a Japanese context, ed. Kunihiro Tetsuya, Inoue Fumio, and Daniel Long, 183-206. Berlin and New York: Mouton de Gruyter. Shibatani, Masayoshi. 1990. The languages of Japan. Cambridge: Cambridge University Press. 2009. Elements of complex structures, where relativizationisn 't: The case ofrelativization. In Syntactic complexity: Diachrony, acquisition, neuro-cognition, evolution, ed. Talmy Givon and Masayoshi Shibatani, 163-98. Amsterdam: John Benjarnins. Shimabukuro, Moriyo. 2007. The accentual history of the Japanese and Ryukyuan language: A reconstruction. Folkestone: Global Oriental. Shinzato, Rumiko. 1991. Where do temporality, evidentiality and episternicity meet? A comparison of Old Japanese -ki and -keri with Turkish -di and -m4. Gengo Kenkyi199, 25-57. ShOgakukan (ed.). 2000-2. Nihon kokugo daijiten, 2nd edn. Tokyo: Shogakukan. Siebold, Philipp Franz von. 1826. Epitome linguaejaponicae. Batavia. Sohn, Ho-min. 1999. The Korean language. Cambridge: Cambridge University Press. Sorace, Antonella. 2000. Gradients in auxiliary selection with intransitive verbs. Language 76, 859-90.

422

References

Spear, Richard L. 1975. Diego Collado's grammar of the Japanese language: Edited and translated by Richard L. Spear. Kansas: Center for East Asian Studies, The University of Kansas. Sugahara, Norio. 2000. (Kirishitan shiryo o shiten to suru) Chiisei kokugo no kenkyii. Tokyo: Musashino Shoin. Sugimoto, Tsutomu. 1985. Edo no onnakotoba: Asobase to arinsu to. Tokyo: Sotakusha. 1988. Tokyogo no rekishi. Tokyo: Chilo Koronsha. 1989. Seiyojin no nihongo hakken: Gaikokujin no nihongo kenkyiishi (The discovery of the Japanese language by western people: A historical survey of learning and study ofJapanese}. Tokyo: Sotakusha. Suzuki, Isao. 1978. Bongokigen no Nihongoi. Gekkan Gengo 712, 30-5. Takeuchi, Lone. 1987. A study of Classical Japanese tense and aspect. Copenhagen: Akademisk Forlag. 1999. The structure and history of Japanese: From Yamatokotoba to Nihongo. London: Longman. Thunberg, Carl Peter. 1792. Observationes in linguam japonicam. Nova Acta Regiae Societatis Scientiarium Upsaliensis 2/5, 258-73. Tsukishima, Hiroshi. 1969. Heian-jidaigoshinron. Tokyo: Tokyo Daigaku Shuppankai. 1977. Kokugo no rekishi. Tokyo: Tokyo Daigaku Shuppankai. 1986. Heian jidai kunten-bon ronko: okototen-zu kanajitai-hyo. Tokyo: Kyilko Shoin. 1987. Heianjidai no kokugo. Tokyo: Tokyodo Shuppan. (ed.). 1993. Studies in Sino-Japanese (Acta Asiatica 65). Tokyo: ToM Gakkai. Tsunoda, Ryilsaku and L. Carrington Goodrich. 1951. Japan in the Chinese dynastic histories: Later Han through Ming dynasties (Perkins Asiatic monographs 2). South Pasadena: P. D. and I. Perkins. Twine, Nanette. 1991. Language and the modern state: The refonn ofwritten Japanese. London: Routledge. Umegaki, Minoru. 1978. Nihon gairaigo no kenkyii. Tokyo: Kenkyusha. Unger, J. Marshall. 1977/1993. Studies in early Japanese morphophonemics: Yale University dissertation, reproduced by IULC. 1988. Middle Chinese syllable-final stops in Sino-Japanese: A critique of Vance's theory Journal of the American Oriental Society 108/4, 627-31. 2001. Layers of words and volcanic ash in Japan and Korea. Journal of Japanese Studies 27/1, 81-111. 2003. Substratum and adstratum in prehistoric Japanese. In Language contacts in prehistory: Studies in stratigraphy, ed. Henning Andersen, 241-58. John Benjamins. 2007. Dating the collapse of koo-otu distinctions and the status of !mol. In Current issues in the history and structure ofJapanese, ed. Bjarke Frellesvig, Masayoshi Shibatani and John Charles Smith, 339-50. Tokyo: Kurosio Publishers. Unger, J. Marshall and YokO Ito Tomita. 1983. The classification of Old Japanese adjectives. Papers in East Asian Languages and Literatures 1, 52-65. Uwano, Zendo. 1977. Nihongo no akusento. In Iwanami koza nihongo, ed. Ohno Susumu and Shibata Takeshi, vol. V (On'in), 283-321. Tokyo: Iwanami Shoten. Vance, Timothy J. 1986. The development of Chinese syllable-final stops in SinoJapanese. Papers in East Asian Languages 3, 48--62. 1987. An introduction to Japanese phonology. Albany: SUNY Press.

References

423

2007. Have we learned anything about rendaku that Lyman didn't already know? In Current issues in the history and structure of Japanese, ed. Bjarke Frellesvig, Masayoshi Shibatani and John Charles Smith, 153-70. Tokyo: Kurosio Publishers. Vovin, Alexander. 1993. Long vowels in proto-Japanese. Journal of East Asian Linguistics 2, 125-34. 1994. Genetic affiliation of Japanese and methodology of linguistic comparison. Journal de Ia societefinno-ougrienne 85, 241-56. 1997. On the syntactic typology of Old Japanese. Journal ofEast Asian Linguistics 6, 273-90. 2003. A reference grammar ofClassical Japanese prose. London: RoutledgeCurzon. 2005. A descriptive and comparative grammar of western Old Japanese, part 1: Phonology, script, lexicon and nominals. Folkestone: Global Oriental. 2008. A descriptive and comparative grammar of western Old Japanese, part 2: Ar:fjectives, verbs, adverbs, conjunctions, particles, postpositions. Folkestone: Global Oriental. 2009. Man 'yi5shii to Fudoki ni mirareru fushigi-na kotoba to ji5dai nihon-retti5 ni okern ainugo no bunpu {Strange words in the Man 'yi5shu and the Fudoki and the distribution oftheAinu language in the Japanese islands in prehistory). Kyoto: Kokusai Nihonbunka Kenkyusenta. Washio, RyUichi. 2002. Jodai nihongo ni okeru jodoshisentaku no mondai: seioshogo to no hikaku kara rniete kuru mono (Auxiliary selection in Old Japanese: a comparative perspective). NihongoBunpi5/Journa/ ofJapanese Grammar211, 109-31. Watanabe, Akira. 2002. The loss of overt Wh-movement in Old Japanese. In Syntactic effects of morphological change, ed. David Lightfoot, 179-95. Oxford: Oxford University Press. Wenck, Giinther. 1959. Japanische Phonetik, vol. IV Wiesbaden: Otto Harrassowitz. Whitman, John. 1985. The phonological basis for the comparison of Japanese and Korean. Doctoral dissertation, Harvard University. 1997. Kakarimusubi from a comparative perspective. InJapanese!KoreanLinguistics 6, ed. Ho-rnin Sohn and John Haig, 161-78. 1999. Personal pronoun shift in Japanese: A case study in lexical change and point of view. In Function and structure, ed. Akio Kamio and Ken-Ichi Takami, 357-86. Amsterdam: John Benjamins. 2008. The source of the bigrade conjugation and stem shape in pre-Old Japanese. In Proto-Japanese: Issues and prospects, ed. Bjarke Frellesvig and John Whitman, 159-73. Amsterdam: John Benjarnins. Wrona, Janick. 2008. The Old Japanese complement system.: A synchronic and diachronic study. Folkestone: Global Oriental. Forthcoming. The early history of no as a norninalizer. In Studies in Japanese and Korean linguistics, ed. Bjarke Frellesvig, Jieun Kiaer and Janick Wrona, forthcorning. London: Lincom. Wrona, Janick and Bjarke Frellesvig. 2010. The Old Japanese case system: The function of wo. Japanese/Korean Linguistics 17, 565-79. Yamada, Yoshio. 1913/1952. Heian-chi5 bunpi5shi. Tokyo: Hobunkan. 1913/1954. Nara-chi5 Bunpi5-shi. Tokyo: Hobunkan. 1935. Kanbun no kundoku niyorite tsutaeraretarn gohi5. Tokyo: Hobunkan. Yamaguchi, Akiho and Akimoto Morihide (eds.). 2001. Nihongo bunpi5 daijiten. Tokyo: Meiji Shoin.

424

References

Yamamoto, Matsuo. 1955. Keiy0shi ku-katsoyo shiku-katsuyo no imijo no sOi ni tsuite. Kokugogaku 23, 71-5. Yamauchi, Yoichiro. 2003. Katsuyi5 to katsuyokei no tsiijiteki kenkyU. Osaka: Seibundo Shuppan. Yamazaki, Kaoru. 1992. Keiyi5shijodi5shi no kenkyU. Osaka: Izumi Shoin. Yanagida, Yuko and John Whitman 2009a. Aligmnent and word order in Old Japanese. Journal ofEastAsianLinguistics 18, 101-44. Yanagida, Yuko and John Whitman 2009b. NinshO to katsukaku ruikei: Jodai nihongo no daimeishi taikei no shiten kara. Yoshida, Kanehiko, Tsukishima Hiroshi, lshizuka Harumichi and Tsukimoto Masayuki (eds.). 2001. Kuntengojiten. Tokyodo Shuppan. Yuzawa, Kokichiro. 1928. Amakusa-bon Heike monogatari no goho. Kyoiku (Special extraordinary issue), 55-126. 1929/1970. Muromachijidai gengo no kenkyU: shomono no gohi5. Tokyo: Kazama Shobo. 1957. Edo-kotoba no kenkyU. Tokyo: Meiji Shoin. 1970. Tokugawa-jidai gengo no kenkyU. Tokyo: Kazama Shobo. Yuzawa, Tadayuki. 1996. Nihon kanjion-shi ronki5. Tokyo: Benseisha.

Index of main grammatical forms

a, etc.) demonstrative 246-7, 353 a, are 1st person pronoun 136, 138 -(a)ba conditional 54, 57, 65, 70, 72, 77, 121,227,292-3,327 -(a)baya optative 227, 240, 242, 326 -(a)de negative gerund 69-71, 270, 236 -(a)hen negative 402 -(a)ide negative gerund 336 -(a)m-, -(a)il-, -(a)u- conjectural 55, 62, 64-6,70,72,77-8,82,121,236,240-2, 330-2,334,341 -(a)maposi, -(a)maosi- optative 236, 241-2, 338 -(a)masi subjunctive 64-5, 70, 78, 236, 330 -(a)mau- negative optative 236, 241-2 -(a)n- negative 69-71, 112-13, 121, 330, 336 -(a)na- negative 398, 401 -(a)na optative 56, 110, 112, 121, 227 -(a)namu optative 56, 227, 242, 326 -(a)nandar-, -(a)nanda, -(a)nande negative past tense 3 36-7 -(a)nap-, -(a)nana, -(a)nani EOJ negatives 154, 401 anala 'you' 247 -(a)ne optative 56, 108-9 -(a)ni negative 69-71 *ani negative 120-2 -(a)nze-, -(a)iize- intentional 241, 329-30, 332, 334, 341 -(a)p- activity verb derivative 52 -ap- reciprocal 52, 79 ar-, grammatical uses 40, 48, 68-9, 70, 79, 90, 95,102-5,152,233-5,238-9,327, 339-40,343-4,350-2,353,395,400, 401-2 -(a)r- detransitivizer 52, 238 -(a)re- passive 63, 112, 121, 236 arinsu 'exist' (polite) 279 -(a)s- respect 61, 110, 112, 121, 236, 371 -(a)s- transitivizer 52 a(-) (anala, ano, are, asoko, ali,

-(a)sime- causative 61, 63, 112, 121, 236-7,

371 -(a)u volitional 330-2, 334, 349 -(a)il-, -(a)u- conjectural 240, 330, 333-4 -(a)iize-, -(a)uzuru intentional 241, 329-30,

332, 341 -(a)ye- passive 61, 63, 112, 121, 236, 273 -(a)zar- negative 40, 236, 336-7 -(a)zi negative conjectural 55-6, 93, 112,

121,242,326 -(a)zu negative 40, 61, 64, 69-71, 93, 112,

121, 236, 330 -(a)zunba negative conditional 336-7 bakari 132, 134 be- (bei, bekasi-, hera-, besii-) necessitive

123,234,242,339,397 -bwi- derivative 52, 99 byoo necessitive 339, 350 da copula 235, 341-4, 394-5 da genitive 131 dare 'who' 354 daroo conjectural copula 394-5 datta past copula 395 dattaroo past conjectural copula 395 de copula gerund 233, 243, 328, 372 de ar- copula 235, 341, 343, 400 dea copula 342, 400 des- (desu, deshita, desho etc.) polite copula

334-5, 372, 395 dewa na- negative copula 343, 395 do- (doko, dona/a, dono, dore, doli, do etc.)

interrogative 246-7, 353-4 dya copula 341-2, 344, 395, 400

dyar- copula 341 e allative 243 -e- potential auxiliary verb 52, 79, 338 -e- transitivity switch 52, 119 -(e)ba provisional 57, 82, 112-13, 292-3, 327, 333

425

426

Index of main grammatical forms

-(e)do(mo) concessive 57, 112, 133 -er- stative 236, 239, 330 ga complementizer 128, 356, 367, 368 ga genitive 126-31, 138, 243-4, 354, 366-8 ga nominative 126-31, 354, 366-8 gana desiderative 241-2 gane conjunctional particle 133 gani conjunctional particle 133 -gar- 242 gati '-prone, be likely to' 235, 242 -ge 235, 242, 339 go- honorific 370 gozar- 'exist' (polite) 340, 373 i- (ika(-), ilw-, itu etc.) interrogative 136,

140,142-3,246-7 i- 'exist', see wi-, i- 'sit down' i adjectival copula 340-1, 344, 394 -i imperative 326 i nominative 131-2 -ide- auxiliary verb 52 idoko 'where' 246-7 idu- (iduku, idure, iduti etc.) interrogative 136-7, 140-2 -{i)ge- evidential 338-9 -{i)kem- past conjectural 48, 236, 241-2 -{i)ker- modal past 236, 242, 330-1 -{i)ki- simple past 62, 64-5, 72-9, 96, 109, 111-13, 122, 236, 241, 331 -{i)kyeba simple past conditional 72 -{i)kyeku simple past nominal 62, 72 -{i)kyem- simple past conjectural 65 -{i)kyer- modal past 48, 62, 64-5, 72-9, 102, 111 -{i)kyeraku modal past nominal 62, 75 ima 'now' 142, 235 -{i)marase- polite 338, 373 imas- 'be, come, go' (respect) 371 -{i)mas-, -{i)mase- polite 52, 338, 372-3 -{i)n- perfective 61,64-7,73, 97, lll, 123, 236,240,330 -{i)sar- respect 369 -{i)seba simple past conditional 62, 72, 74, 122 -{i)si- simple past 62, 64-5, 72-9, 96, 109, 111-13, 122, 236, 241, 331 -{i)sika, -{i)sikaba, -(i}sikado simple past 62, 64-5, 72-9, 96, 109, lll-13, 122, 236, 241, 331 -{i)siku simple past nominal 62, 64-5, 72-9, 96, 109, 111-13, 122, 236, 241, 331 -{i)soo-, -{i)sX>- evidential 338, 396 -{i)sooroo polite 338, 372-3, 395 -{i)ta- desiderative 326, 338 -(l)ta past tense 69, 239, 331-2, 347, 350

-{i) tar- stative 104, 23 6, 2 39 -(!)tara conditional 327, 333 -(l)tari representative 330-1 -(l)taroo, -(l)taroo past conjectural 332-5 -{i)te gerund 57, 69, 111, 123, 230 -{i)te- perfective 61, 64, 66-8, ll1, 230, 236, 240,330 -{i)te ar- periphrastic stative 69, 103, 238-9 -{i)tutu continuative 57, 69, 111, 327 ja na- negative copula 343, 394 ka interrogative focus particle 55, 247-252 ka(-) (kano, kare, kasiko) demonstrative 139, 246-7,353 ka, kaku, kakute, kau, koo, ko 'like this' 141, 246-7 kakur- 'die' (respect) 371 kamo final particle 89, 108, 133, 241 kana exclamatory particle 241 kar- adjectival copula 40, 90, 233-4, 339 kara ablative 132, 134 karoo, karoo adjectival copula conjectural 334-5,340-1,394 kasi emphatic 241-2 katta adjectival copula past 340-1, 402 kau 'like this' 353 kere, kereba, keredo adjectival copula 82, 233,340-1 keredomo conjunction 341 ki adjectival copula adnominal 82-3, 232 ko(-) (koko, konata, kono, kore, koti, ko etc.) demonstrative 136, 139-42, 245-7, 353-4 ko-, grammatical uses 52, 67, 108 -kose- auxiliary verb 96, 108-9, 135 koso focus particle 55, 132, 247-51, 255, 328, 358, 361 koo 'like this' 353 leu adjectival copula infinitive 70, 80-2, 85-6, 90, 122, 133, 232-3, 339-41, 343-4, 393-4,397 kunba adjectival copula conditional 340, 341, 393 kyeba, kyedo adjectival copula 81-2, 122 kyeku adjectival copula nominal 81-2, 122 kyem- adjectival copula conjectural 122 kyenaku adjectival copula negative nominal 122 kyere, kyereba, kyeredo adjectival copula 81-2, 122 made allative 243 made{ni) 132-3 -rna/rase- humble 373 -mas- respect 52 masizi- negative potential 93, 123-4

427

Index of main grammatical forms -matur- humble 52 mawir- 'enter' (humble) 338, 371 mawos- 'say, speak' (humble) 52, 193, 371 -mawos- humble 52 mazi- negative necessitive 241-2, 335, 339 -mek- derivative 242 mer- evidential 102, 232, 241-2, 339 mi adjectival copula infinitive 86-9, 117,

232 mi- honorific 370 miyu final particle 86, 133, 135, 241 mo emphatic topic 132 moga(mo) final particle 133 mogana optative 242 mono 93,132-3,268,292-3 mono-kara (ni) conjunctional particle 133-4 mono-wo conjunctional particle 133-4 mono-yuwe (ni) conjunctional particle 133,

134

pa topic 132, 223, 249-51, 292-3 paber-,pabyer- humble, polite 372 por- 'wish, want' 88 ppoi evidential 396 puser- 'be lying down' 352 pye,pe, we, e allative 126, 132, 135-6,

243 ram- present conjectural 107, 123-4, 242,

334,339 -rare- passive 236-8, 337-8, 371 rasi- evidential, derivative 395-6 rasi- presumptive 123-4, 241-2 -re- potential 338 -ro imperative 133, 154 ro interjectional particle 133

roo conjectural 334-5, 341 sa adjectival copula exclamatory 81-2, 84-5,

117, 122, 232 na (-no) adnominal copula 135, 338-9, 342,

-sabwi- derivative 52, 99

395-6 nagenitive 126,131,242 na prohibitive 56, 133, 326, 329 na, nare 2nd person pronoun 136-7, 245 na- ... -so prohibitive 56, 108, 120, 227, 326 nado 245, 359 namo, namu focus particle 55, 132, 247-52, 254-5,358-9 nape (ni) conjunctional particle 133 nar- copula 95-6, 104-5, 124, 233-5, 339, 341-3 nar- evidential 123-4, 242, 339 nara conditional 327 nara(ba) copula provisional 395 nasu, nosu semblative 122-3 ni conjunctional particle 78, 133-4, 244 ni copula infinitive 94-5, 134-5, 233, 235, 243,287,339,342-4,395 ni dative 126, 243 nile copula 93-4, 135, 233, 243, 328, 339, 341,343 no copula 94-5, 126-8, 135, 233, 243, 342 no genitive 126-31, 137-8, 243-4, 366-8 no nominalizer 127, 354, 364-5, 368, 126 nosu semblative 122-3

-sai imperative 369

o accusative 126, 129-31, 132, 135, 223, 243 o-, on-, oon-, opomi- honorific 370 odyar- 'exist' (polite) 350,373 on-, oon- honorific 370 ono, onore reflexive pronoun 136, 142, 246 opomi- honorific 370 or- 'exist', see wor-, or- 'be sitting; exist' oryar- 'exist' (polite) 350, 373

-sase- causative 63, 236-8, 337 sate, sau 'like that' 141, 353 -sauraw- polite 372 se-, grammatical uses 67, 108, 229, 287 shi conjunctional particle 395 si 3rd person pronoun 136, 138-9 si adjectival copula conclusive 81-4, 122,

233 sika 'like that' 141, 246, 271 so(-) (solw, sonata, sono, sore, soli, sao, so

etc.) demonstrative 136, 139-42, 245-7, 353-4 so, zo focus 55, 132, 247-51, 253-4, 354, 358-61, 365 soo-da 'hearsay' 338, 396 -sooroo, -sooraw-, -sooroo, -sao polite 372-3, 393 sao 'like that' 320, 353 Ia, tare 'who' 139, 246, 353-4 -tamap- respect 52, 371 tar- copula 40, 104-5, 233-5, 339, 341 -tatematur- humble 371 tater- 'be standing' 351-2 to comitative 126, 223 to complementizer 134 to conjunctional particle 53, 55, 78, 133, 243,

244-5 to copula infinitive 93-4, 134-5, 233, 243,

245,342-3,395 to(mo) conjunctional particle 53, 133 tolwro-no 274, 358 tote 245 tu copula 80, 93-4, 135

428

Index of main grammatical forms

tu genitive 126, 131, 135, 242 tukapematur- 'serve' (humble) 371

ya interrogative focus particle 55, 247-53,

354-4 yaroo, yarooparticle, copula 335, 359, 394-5

u adjectival copula infinitive 232-3, 340-1,

348-50, 385-6, 397-8 wa topic 132, 223, 249-51, 292-3 wa, ware 1st person pronoun 136-9, 245-7,

353 we interjectional particle 133 wi-, i- 'sit down; exist' 69, 98, 101-2, 106,

135, 240, 350-2 witar- 'be sitting' 240, 351 wo accusative 126, 129-31, 132, 135, 223,

243 wo conjunctional particle 55, 126, 133, 135,

244 wo interjectional particle 126, 133, 135 wo-ba accusative+ topic particle 126, 132 wor-, or- 'be sitting; exist' (see also or-) 69,

101-2,135,240,350-2

-yail '-like' 235, 242 yail, yau- evidential 235, 288

-yer- stative 48, 61, 64, 68-9, 79, 113, 240, 331 yo emphatic 107, 114, 125, 133, 242, 356(n.), 402 -yo imperative 107, 114, 227, 326 yo interjectional particle 133 -(y)oo volitional 330 yoo-, yro- evidential 339 yoo-ni complementizer, conjunctional particle 339 yori, yuri, yu, ywori, ywo ablative 126, 134, 223,243 yuk-, grammatical uses 67 yuwe (ni) conjunctional particle 133-4 zo focus 132, 247-51, 253-4, 354, 358-61,

365 ya copula 359, 395 ya interjectional particle 133

General index

a- stem 5, 111-12, 114, 120-1,401-2 absolute respect 371 Abutsu 298 Account of the dwarfs 20-1 accusative 126, 129-31, 243 accusative-with-infinitive construction 95 adjectival copula (see also Index of main grammatical forms: ar-, grammatical uses, copula) 80-93, 104, 117, 120-2, 233-5,339-41,343-4,393-5 adjectival nouns 235, 288 adjectives (see also adjectival copula) 79-93, 119, 224, 235, 288 jiku 93, 123 shiku 81, 83, 90-3 adnominal clauses 53-5, 58, 80, 94, 127-8, 245,274,292-3,355-8,362,365, 366-7 adnominal form (see also merger of conclusive and adnominal forms) 53, 83-4,94, 100-1,113,115-17,152-3, 224,227,232,244,247-9,256-7,336, 353, 362, 365 Ainu 37, 145-6, 154 Akichi 11 alphabet, see Iroha-uta Ame-tsuchi no kotoba 165-8 analytic forms, see Index of main grammatical forms: ar-, grammatical uses anusvara 164 apophonic alternations 45-6, 119, 131 ar- extended forms, see Index of main grammatical forms: ar-, grammatical uses archival records, see komonjo Arisaka's Law 44 Ars grammaticae iaponicae linguae 298, 301-2 Arle breve da lingoa Iapoa 298 Arle da lingoa de Iapam 6, 57, 279, 298, 301-2

aspect 53, 59,64-9,238-240,327,331 assibilation 322 athematic verbs 119-20 auxiliaries 58-9, 332 auxiliary verbs 52, 67, 108, 338, 371-3 azuma uta 24, 28, 151 basic paradigms 116, 344 bigrade theme 118-9 bigradeverbs 96-7,116,118-20,346-7, 391-3 borrowing, see grammatical borrowing, loanwords bound moras 169-70, 184-91, 200, 304, 309-10 Bungofudoh 23,25 Bussokuseh-ka 23, 25, 104-5 ICI, see bound moras Calepino 302 case marking, object and subject 127-31, 243,366-8,410-11 case particles 94, 125-32, 134-6, 242-4, 269, 271, 410-11 causative 63-4, 236-8, 33 7, 346 Chinese (see also Japano-Chinese, SinoJapanese) 2, 11-12, 17, 21, 23-5, 30-1, 162, 164, 170, 179, 181-3, 185, 199-200,258-9,261-80,309,317, 357,370 as a reading language in Japan, see Japano-Chinese loanwords 144-50, 284-7, 290-4 script, adaptation to write Japanese 12-16, 17-18,158,263-6,268 Christian materials (hrishitan shiry6) 2, 299-303 Classical Japanese 2, 116, 158, 179, 302, 381 clefts 55, 248, 251, 293, 3 62 Collado, Diego 301-4, 312, 323 common language 336, 3 77-9, 380, 383, 395, 397-8, 401

429

430

General index

complement clauses 58, 87, 227, 241, 362-5 complementizer 75, 125, 128-9, 134, 245, 260,273,274,339,355-8,364-7 concessive 57, 82, 112, 270, 341, 344, 393 conjunctional particle, see Index of main grammatical forms: to(mo) conclusive form (see also merger of conclusive and adnominal forms) 53-5, 83-4,100,107,152-3,224-6,249 as stem 101, 391-2 conditional 57, 65, 77, 82, 86, 121, 152, 227, 292-3, 327, 330, 333, 341, 393, 398, 402 conjectural (see also intentional, negative conjectural, past conjectural, present conjectural, volitional) 64-5, 78, 82, 88, 112, 121, 240-2, 330-2, 334-5, 341,393,394 conjugation classes (see also verb classes) 96 conjunctional particles 133, 134-5, 244-5, 271,339,341,366,395 consonant base verbs 96-7, 101-16, 113 continuative form 57, 69, 123, 327 continuousness variation 36-7, 196, 202-3, 205 contraction of contiguous vowels, see vowel contraction copula (see also adjectival copula, Index of main grammatical forms: or-, grammatical uses) 94-6, 117, 135, 232-5,339-44,393-5,400 as source of other grammatical morphemes 121-3, 135 covert form (hifukukei) 45-6, 119 ICy-/ 31-2, 184-6, 199-200, 319, 322-3, 325 da Silva, D. 302 Daijionji sanzohoshi-den 180, 182, 286-7 daku'on, see mediae dakuten (see also handakuten) 160, 163-5, 170,175 dative 41, 126, 135, 243 demonstratives, see pronouns, proto-Japanese demonstratives derivation (see also bigrade theme) 52, 81, 87, 91, 93, 141, 229-30, 232, 235, 242, 288,338-9,396 desiderative 88-9, 230, 241, 338 dialect eradication 212, 380-1 Dictionorium Latino-Lusitanicum ac Japonicum 298, 302 distal demonstrative, see pronouns Dizang shilim jing 160 Doctrina Chris/an 302

I.e/, see onglides Early Middle Chinese 15-16, 30-1, 35, 37, 147-50, 185, 199-200, 204, 277-8, 281-5,290,317 and Sino-Japanese 282-4 eastern dialect features in the standard language 361, 395, 397-402 Eastern Old Japanese 123, 138-9, 146, 151-4,400-2 Edo Japanese 378-9, 382-3, 398 Engishiki Norito, see Norito Ennin 37, 183 Epitome linguae japonicae 382 equivalence, in writing 17-18, 163, 165, 171-3 Esopono fahulas 298, 301 etymology of particles 134-6 etymology of verb suffixes 120-3 evidential 72, 76, 86, 102, 123-4, 133, 232, 241-2,338-9,395-6 exaltation 63, 369-71 exclamative sentences 55, 58, 83, 251, 253, 256-7,329,354-5,358-61 exclamatory 5, 55, 82, 84-5, 100-1, 115, 232,247-9,253,255,256,344,358 as stem 112-13 loss of 328 existential verbs (see also Index of main grammatical forms: or-) 69, 90, 94-6, 103,233,339-40,343-4,350-2,366, 372-3,402 exposed form (roshutsukei) 45, 119 extensions 107, 123-4, 153, 232, 241-2, 334-5,339,395-6 Iff 165, 201-2, 204-5, 311-13, 315-16, 386-7 fli.nqie 177, 183,277 Feiqe monogatori (see also Heike monogatari) 298, 301 Fernandez, J. 302 final particles 133, 241-2, 354, 358, 359, 365 finite verb forms 53 focus 132,248-51,359-61 particles (see also kakari particles) 132 Francis Xavier 299 Fray Luis de Granada 302 fricative variants of obstruents 36-7, 196, 202-3, 205 fricativization of /p/ 36-7, 311-16 fudoki 25 Fujiwara no Hamanari 25 Fujiwara no Tamefusa's wife 174 Fujiwara no Teika 174 Furukawa Koshiiken 377 Futabatei Shimei 379

General index gairaigo 403--4 gapless adnominal clauses 356-7 genbun'itchi 379,411 genitive particles 41, 123, 126-8, 131, 223, 242--4,355-8,366-8 differences between no and ga 126-9, 367-8 genitive subject marking 127-31, 243, 366-8 Genji monogalari 175-6, 179-80, 200, 241, 286-7 Genpei josui/d 297-8 gerund 57, 71, 86, 88, 111, 123, 189, 226, 327-8,347,400 Gishi wajinden 20 Giya do pekadoru 302, 316 glossaries 182, 277 godan ('quinquigrade') verbs 392 gojuonzu 115,158,168,177,182 go-on 275-83,290-1,317,409 grammatical borrowing 2 70--4, 41 0-11 Granada, Fray Luis de 302 Gulingyun 278 Guia de pecadores 3 02 Gukansho 298 lgw-1, see lkw-1 Gyoa 174 Han'on saho 180, 183 handakuten 165, 302-3 Harimafudo/d 23, 25 Hashimoto Shinkichi 28-9 Hashimoto Shiro 139--40 headless nominalization 55, 83, 245, 362--4 heavy syllables 184-6 Heiji monogatari 297-8 Heike monogalari (see also Feiqe monogatari) 176, 297-8, 372 hentai kanbun 13, 266-8 hentaigana 158, 160, 162, 170, 317-18 hifukukei, see covert form hiragana 158-62, 169 hiragana literature 179-81, 287 historical kana-spelling 4, 173-7, 380, 385 Hilachifudo/d 23, 25, 151 Hizenfudo/d 23, 25 Hogen monogatari 297-8 Hojo/d 176, 297-8 Hokke mongu 180, 182, 261 honorific language (keigo) 79, 139, 267, 369-73, 378 humble language (kenjogo) 52, 350, 369-73 hyangch 'al script 13 hybrid words 403-4

Ill, Iff, see bound moras, monophthongization [l'], see nasalization

431 ichidan verbs, see monograde verbs idu script 13, 16 Imbe no Hironari 25 imperative 3, 55, 101, 114, 115-17, 125, 133, 154,227,242,326,369,402 infinitive 5, 56-7, 68-9, 82, 85-90, 94-5, 101, 103, 108, 114, 115-17, 120-3, 194-5,224-6,240,327,350,370,402 as stem 109-11, 230-1, 232 intentional 236, 241-2, 328-30, 332--4, 341, 390,397 interjectional particles 133 internally headed relative clause, see headless nominalization interrogative pronouns, see pronouns interrogative sentences, see questions Irohajiruisho 168, 180, 183 Iroha-uta 165, 167-8, 173, 182-3,206 Irop 'a 298-9, 322 Ise monogatari 176, 179-80, 204, 286 Ishizuka Tatsumaro 29 Izayoi nik/d 298 izenkei (see also exclamatory) 5, 115-17 Izumo fudo/d 23, 25

Japano-Chinese (see also Sino-Japanese) 11, 16, 183, 200, 274-80, 282, 284, 286-7, 290-1, 293 phonetics and phonology 16, 277 Pe], see onglides Jesuits 57, 299-303, 404 Jien 298 Jik/dnsho 297-8 jiku adjectives 93, 123 Jinno shoto/d 298 Jizo-jurin-gy6 160 Jogon 178 k-irregular verbs (see also Index of main grammatical forms: /w-, grammatical uses) 97, 107-8, 116, 120, 393 kaeriten 259-60, 262 Kagero nik/d 176, 180 ka(gyt))-hen(kaku) verbs, seek-irregular verbs Kaifiiso 11 kakari particles 247-9, 251-5, 358-61 kakari-musubi 247-57, 328-9, 358-61, 365 Kakyo hy6shi/d 23, 25, 137 Kamo no ChOmei 297 kan-on 275-83,290-1,317,409 kan'yt)-on 280-1,283, 291,317 kanbun (see also hentai kanbun) 258, 260, 266-8 kanbun-kundoku 11, 87,131,158,181,227, 234,237,242,258-74,292,300,341, 357,358,410

432

General index

Kanchi 178 kanji-kana majiribun 157-8, 162, 181, 237, 266,279,287,290,297 Kansai dialects 214-16, 236, 349, 350-2, 381,393-5,397-8,400-2 Kanto dialects 361, 383, 397-402 katakana 158-62, 169, 259, 263, 404, 409, 411 katsu)IOkei system (see also basic paradigms) 114-18, 344, 392 Keichil 29, 175, 178 keigo, see honorific language keiyodoshi, see adjectival nouns kenjogo, see humble language Kino Tsurayuki 157,179-80 Kindaichi Haruhiko 212, 219 Kindaichi's Law 219 Kinsui Satoshi 101, 135, 245, 351, 365, 373,410 kirishitan shiry6, see Christian materials Kitabatake Chikafusa 298 ko-rui and otsu-rui syllables 26-34, 44-50, 205-6 Kogoshiii 23, 25 Knhon setsuwa-shii 180-1 Kojiki 11, 13, 18-19, 23-4, 28-30, 34, 72, 104, 138, 153, 264-7 Kokin wakashii 179-80, 286 lwkugo (national language) 380 Kokugo-chosa-iinkai 381 lwkuji 268 lwmonjo 22, 29 Knnjaku monogatari-shii 180-1 Knnkomyo saishoo kyo 182 Konkom)IO saishoo kyo ongi 167, 182 Korean loanwords 146-7 Korean script 13, 16 ku adjectives 81, 90-2 Kilkai 160, 164, 168 kundoku, see kanbun-kundoku kunten 158, 160, 181-2, 187, 192, 214, 259-63, 266, 268, 269 kuntengo 268-74, 411 Kusakado Nobutaka 29 lkw-1 31-2, 184-6, 199-200, 227-8, 304-5, 310-11, 325, 387 Kyoto dialect 210, 214, 217-20, 349, 378, 397-8,400-2

loanwords (see also Sino-Japanese loanwords) 144-50, 302, 316, 325, 386-9,403-12

phonological adaptation of 150, 282, 406

logographic writing 4, 12-14, 157-8, 181, 259,263-8,287,291-2

long syllables 184-6 long vowels (see also monophthongization of NU, VII) 216, 309, 319-21, 323

in Sino-Japanese 321-2 lower bigrade verbs 96-7, 116, 120, 346-7, 391-3

lower monograde verbs 227-9, 355, 391-3 Lu Fayan 278 Lyman, Benjamin Smith 41 Lyman's Law 41 1-ml versus /-n/ 184, 200 Makura no soshi 175-6, 179-80 man'yogana 14-19,27, 28-9, 152, 157-8, 160, 162-5, 169-70, 183 Man'yoshii 14, 17-19, 23-4, 28-9, 72, 82, 151,266,270,284-6 Martin, Samuel E. 32, 44, 49-50, 210, 217, 220,222,224,302 Masukagami 297-8 Maximus Planudes 301 mediae (see also prenasalization) 34-6, 38, 40,42-3, 163-5, 184-5, 188-91, 199, 214,306-8,389 medial voicing 34-6, 189-91, 196, 203, 306-8 medium voice 63 Meigaku 177-8, 183, 194 meireikei (see also imperative) 115-17

merger of conclusive and adnominal forms 328-30, 332, 340, 354-65, 392

mesial demonstratives, see pronouns mid vowel raising 47, 50, 142, 153 Miller, Roy Andrew, see Account of the dwaifs

mimetics 315-16, 403 Minamoto no Tamenori 167 Minamoto no Tomoyuki (Gyoa) 174 Miyake, Marc Hideo 16, 31, 37, 147 mizenkei (see also a- stem) 5, 111, 115, 117, 118

modal past 59, 64-5, 72-8, 102, 121-2, 236, labial onglide, see onglides labialization before /u/ 38, 196-7, 388 light verbs (see also Index of main grammatical forms: ar-, grammatical

uses) 67, 108, 229, 287, 401-2 liquid (/r/), distribution of 34, 43, 199 loan translations 148, 270-2, 285, 408-10

242,331

modality 53, 59, 64-5, 123-4, 133, 240-2, 271 Mogyu-sho 298-9 mokkan 16, 22, 160 monograde verbs 106, 115-16, 228-9, 391-3

General index monophthongization of NV, VI/ 200, 209, 319-22,324-5,348-50,380,386 moraic phonemes, see bound moras Moshi-sho 298-9 Moto'ori Norinaga 178, 249, 280, 357 Murasaki Shikibu 179 IN/, see bound moras n-irregular verbs 97, 105-6, 116, 119-20, 392-3 na(gy6)-hen(kaku) verbs, seen-irregular verbs nasality (see also prenasalization) 184, 187-91,197-8,306-10,347 harmony 190-1, 307 nasalization 38, 307, 309, 387 necessitive 123-4, 241-2, 339, 397 negative auxiliary 64-5, 69-72, 103-4, 120-1,233-4,336-7,397,401-2 negative conjectural 55-6, 61, 112, 121, 242, 326, 329, 335 negative past 336-7 negative rhetorical questions 70 nidan verbs, see bigrade verbs Nihon shoki 11, 23-4, 28-9, 72, 104, 138, 215 nominal form 55, 57-8, 75, 83-4, 88, 113, 117,227,232,273,300 nominalization (see also headless nominalization, nominal form, nominalizing no) 57, 84, 87, 111, 131, 227,243,244,256,272-3 nominalizing no 127, 362-5, 367 nominative case particles 126-7, 130-1, 242, 243,354,366-7 non-finite verb forms 53 non-proximal demonstratives, see pronouns nonpast 332-3, 363, 392 Norito 16, 18-19, 20, 23-4, 77, 266, 270 numerals 285, 289, 291 1-rjl, Early Middle Chinese 164, 170, 199,

282-3,290,309 [rj], [N] in NJ 307, 309, 318, 388 /.o/, see onglides 0 no Yasumaro 13 Observationes in Linguam Japonicam 382 Ochikubo monogatari 176, 180 Ohno Susumu 32, 119 Okagami 180 okoto-ten 260, 262 Old Chinese 21, 147, 148-50 onbin 6, 35, 191-9, 230-4, 240-1, 330, 334, 344,347-9,386,394,398-400 ondoku 11,258,274-92

433 ongana 14-16,19,35,37,204,277-8 ongi, see glossaries onglides, labial and palatal 47, 206-10, 283, 310,319-21,323-5,387 onzu, see gojiionzu, sound tables optative 56, 108, 112, 121, 133, 227, 240-2, 326 orthographic norms, see historical kana-spelling orthographic overdifferentiation 268 otsu-rui, see ko-rui and otsu-rui syllables Otsuki Fumihiko 177, 381 Oya Torn 28

/oo/ 305-6, 319-21, 385-6 !pi 35, 37-8, 165, 171-2, 178, 302-3 alternations arising from changes of 41, 313-15 changes affecting 201-5, 311-16 in mimetics 315-16 Paekche 11, 13, 146-7, 275 palatal onglide, see onglides palatalization before /i, e/ 38, 64, 196-7, 206, 322-3,387,388,397 passive 63-4, 236-8, 273, 337-8, 346, 410 past conjectural 236, 241-2, 330-2, 334-5, 336,354,390,393,394 past provisional 327, 336 past tense (see also modal past, negative past, past conjectural, recent past, simple past) 64-5,72-8,327,330-3,336 perfect 68, 74-6, 239-240, 330-1, 353 perfective auxiliary 64-8, 72, 76, 105, 123, 238-40, 331 periodization of Japanese 1-2 personal pronouns, see pronouns phonographic writing 4, 12-16, 20-1, 145, 157, 263-6 pitch, notation of 163-4, 174, 214-15 pitch accent (see also prosody) 210-11 polite language 370, 372-3, 395 polyvalence, in script 17-18, 172 postnasal neutralization 189-91, 307-9, 314-15, 347 potential 63, 79, 93, 123-4, 338 prefixes 56, 79, 288, 313 prefixes, honorific 267, 370 prenasalization 4, 34-6, 164-5, 189-91, 197-8,306-7 present conjectural 123-4, 242, 334, 339 presumptive 123-4, 241-2 presupposition, see focus primary verbs, see athematic verbs progressive 69, 240, 327, 351-2

434

General index

prohibitive 56, 79, 108, 120, 133, 227, 326, 329 pronominal use of genitive (see also nominalizing no) 243-4 pronouns, demonstrative 121, 128, 136-7, 139-43, 246-7, 353-4 interrogative 136, 140-3, 246-7, 353-4 personal 128, 136-9, 142, 243, 245-6, 353-4,366 pre-Old Japanese 142-3 reflexive 136-8, 142, 246 short and long forms 136-8, 247 prosodic length 224 prosodic shapes 211-12 of morphologically complex forms 219, 222-6 prosodic systems 210-12, 220-1 prosodic word classes 212-19 prosody 210-26,398 proto-Japanese demonstratives 142-3 proto-Japanese morphology 118-23, 153 proto-Japanese phonology 31,41-50, 212-13, 215, 220 provisional 57, 112, 292-3, 327 proximal demonstratives, see pronouns pseudo-cleft constructions 55, 362 purposives 55, 78, 126, 133, 245 /Q/, see bound moras Qieyun 277-8 quadrigrade verbs 96-7, 115, 120, 355, 392-3 questions 55, 70, 83, 132, 252-3, 256-7, 354,359-61,365,366 quotative framing construction 75, 272-3 r-irregular verbs (see also Index of main grammatical forms: ar-, grammatical uses) 97, 101-5, 116, 120, 232, 392-3 ra(gyo)-hen(kaku) verbs, see r-irregular verbs Rakuyoshii 165, 298, 302-3 Ramsey, S. Robert 214 reading tradition 18 rebus writing 18 recent past 67, 240, 331, 334 reflexive pronoun, see pronouns rekishi-teki kana-zukai, see historical kana-spelling relative clauses, see adnominal clauses ren 'yOkei (see also infinitive) 5, 115-17 rendaku 40-1,42-3,163,173,315 renjo 200-1, 318-19 rentaikei (see also adnominal form) 115-17 reported speech, introduction of 75, 272-3

representative form 330-1, 390, 398 respect language (sonkeigo) 52, 62-3, 236-7, 337-8, 369-73 restrictive particles 132-3 rheme 249 rhetorical questions 70, 253, 256 Rodrigues, Joao 6, 279, 301-4, 308-9, 357, 367,369,397,400-1 Rongo-sho 298-9 roshutsukei, see exposed form /-ru/, loss of word final 353 Ruiju-myogi-sho 164, 180, 182, 214-15, 217, 254 s-irregular verbs (see also Index of main grammatical forms: se-, grammatical uses) 97, 107-8, 116, 120, 329, 346, 393 sa(gy6)-hen(kaku) verbs, see s-irregular verbs sakimori uta 24, 28, 151 Sanbo elwtoba 180-1 Sangoshiiki-chii 286, 297-8 Sanskrit studies 183 Sarashina nikki 180 secondary conjugations, see Index of main grammatical forms: ar-, grammatical uses secondary verbs, see bigrade verbs Sei Shonagon 180 sei 'on, see tenues semblative 121-3 Senmyo 16, 23-4, 77-8, 87, 131-2, 254, 266, 270,286 senmyO-gaki 16, 24 serial verb constructions 327-8 setsuwa 181, 270, 287, 297 Setsuyoshii 298, 318 Shasekishii 297-8 Shiganikkai 298-9 Shiki-sho 298-9, 338 shiku adjectives 81, 83, 90-3 Shilla 146 Shinren 183, 205 Shinsenjikyo 180, 182 Shittan kuden 180, 183, 205 Shittan yoketsu 180, 183, 194 shomono 2, 299,308,318 shiishikei (see also conclusive form) 115-17 Shiiyaku-sho 298-9 sibilant obstruent 34-8 Siddham 164, 178, 183 Siebold, Phillipp Franz Balthasar von 382, 386-7 simple past 59,64-5,72-8,96,117,121-2, 236, 331

General index Sino-Japanese (see also Japano-Chinese) 278-84, 290 coinages 291, 408-10 loanwords 170-1, 184-5, 199-200, 229, 235,284-91 phonology 199-201, 278-84, 316-19, 321-2 social deixis 52, 369 Soga monogatari 176, 297-8 sonkeigo, see respect language soro-bun 3 73 sound tables (onzu) 177-8 speaker commitment 76-8 split intransitivity 67-8 standard Japanese 177, 210, 324, 350, 352, 379-81, 382-3, 391, 393-6, 397-402, 406, 410-11 eastern dialect features in 361, 395, 397-402 standardization 212, 3 79-81 stative 59, 64-6, 68-9, 74-6, 79, 102-4, 113, 152,238-240,327,330-1,350-2 stems in verb morphophonology 109-13, 194,344 subject marking 127-31, 243, 366-8,410-11 subjective modality 77 subjunctive 64-6, 78, 121, 330 substantive noun 57, 226 suiryijkei 392 syllable final/-m/, see 1-ml syllable final/-tl, see 1-tl syllable structure 39, 184-99 syllable weight 184-6 1-tl, syllable final 169-70, 200, 283, 316-19, 323 Taiheiki 297-8 Taini-uta 165-8, 206 Taketori monogatari 176, 179-80 Teika, see Fujiwara no Teika teineigo, see polite language tenioha 260 tense 53, 59, 64-5, 72-8, 331-2 tenues 34-6, 163, 184-5, 188-90, 202-3, 214,306-8 terms of address (see also pronouns) 139, 246 thematic verbs, see bigrade verbs theme-rheme 249 Thomas Kempis 300 Thunberg, Carl Peter 382, 386-7, 405-6 10-on 276, 278-81, 290-1 Todaiji fujumonko 158-80 Togashi Hirokage 249 tonality adjustment 38

a

435 tone dots 214-16 topic particles 132 Tosa nikki 157, 175-6, 180,286 transcription conventions 4-5, 32, 191, 198, 384 Tsurayuki, see Ki no Tsurayuki Tsurezuregusa 297-8 Tsutsumi chiinagon monogatari 180

101, lUI, see bound moras [il], see nasalization Ueda Kazutoshi 380, 404 Ujishiii monogatari 297-8 unaccusative 67 unergative 67 Unger, J. Marshall 68, 90, 119, 144-5 upper bigrade verbs 96-7, 116, 120, 391-3 upper monograde verbs 97, 106-7, 115-16, 120,229,355,391-3 Utsuho monogatari 176, 179-80 uvular nasal, 309, 318 verb classes, diachronic classification 119-20 mergers 106, 228-9, 346, 391-3 traditional names 115-16 verbal nouns 229-30, 287 Nil, see monophthongization Vmu no xita nigoru 308-9 Vocabulario da lingoa de lapam 298, 301, 303 vocabulary layers (see also Sino-Japanese) 145, 274, 403-4 voice (see also causative, passive) 63 volitional 330-2, 337, 341, 349, 390-2 Vovin, Alexander 131, 136, 138, 146, 245, 286 vowel base verbs 96-9, 106-9, 113 vowel contraction (see also monophthongization of NV, VII) 44-8, 50, 113, 119-20, 137 vowel deletion 39-40, 99, 120, 323 NUl, see monophthongization

lwl (see also lkw-1) 32, 43, 47-9, 184-5, 199-200, 202-3, 230, 311, 324-5, 389 loss of 205-10, 310-11, 346-7, 387 Wajinden 20 WamyO-ruiju-sho 180, 182 Wani 11 Wei zhi 20 Wenck, GOnther 35, 203 Whitman, John 32, 33, 44, 64, 67, 79, 119, 142 two/ 206-7

436

General index

[wo], see onglides Wo ren chuan 20-1 wooden tablets, see mokkan writing in Japanese, influence from kanbun-kundoku on 158, 263-6, 268 Wrona, Janick 58, 256, 365 Xavier, Francis 299

/y/ 32, 43,47-9, 184-5, 199-200, 324-5

loss of 205-10, 346-7 lye! 206 yodan verbs, see quadrigrade verbs yo'on (see also /Cy-1, lkw-1) 184

Yoshida Kenko 297 Zaitoki 37, 180, 183

A History of the Japanese Language

Read more

A History of Japanese Literature

Read more

A History of Japanese Buddhism

Read more

Japanese Sports: A History

Read more

Language, Ideology and Japanese History Textbooks

Read more

Encyclopedia of Japanese History

Read more

A Companion to Japanese History

Read more

A Cultural History of the English Language

Read more

A History of the English Language

Read more

Towards a History of the Basque Language

Read more

A History of the Greek Language

Read more

A Cultural History of the English Language

Read more

A Brief History of the Spanish Language

Read more

A History of the Korean Language

Read more

A History of the Korean Language

Read more

A Cultural History of the English Language

Read more

A History of the English Language (2006)

Read more

A History of the French Language

Read more

A History of the English Language

Read more

A Cultural History of the English Language

Read more

A History of the English Language (2006)

Read more

A History of the French Language

Read more

Japanese Phrasebook: A Language Survival Kit

Read more

Japanese Phrasebook - A Language Survival Kit

Read more

An Encyclopedia of Japanese History

Read more

Japanese Language Teaching: A Communicative Approach

Read more

History of the French Language

Read more

A History of Language Philosophies (Studies in the History of the Language Sciences)

Read more

The Book Of Five Rings Japanese History

Read more

The Book Of Five Rings Japanese History

Read more

Recommend Documents

A History of the Japanese Language

h \) -o ~ h ' \ J \ I ~ ~\ ( h)) ;, t ~ 1; .\ ) -I tl : l l t A History of the / .., 1r ~ JAPANESE Lan...

A History of Japanese Literature

...

A History of Japanese Buddhism

00 Prelims HJB:Master Testpages HJB 10/10/07 11:14 Page i A HISTORY OF JAPANESE BUDDHISM 00 Prelims HJB:Master Te...

Japanese Sports: A History

Japanese Sports Japanese Sports A History ALLEN GUTTMANN LEE THOMPSON University of Hawai‘i Press ©2001 University...

Language, Ideology and Japanese History Textbooks

1111 2 3 4 5111 6 7 8 9 1011 1 2 3111 4 5 6 7 8 9 20111 1 2 3 4 5111 6 7 8 9 30111 1 2 3 4 5 6 7 8 9 40 41111 Language...

Encyclopedia of Japanese History

CLICK HERE TO VISIT An Encyclopedia of Japanese History compiled by Chris Spackman Copyright Notice Copyright © 2002...

A Companion to Japanese History

A Companion to Japanese History BLACKWELL COMPANIONS TO WORLD HISTORY This series provides sophisticated and authorit...

A Cultural History of the English Language

A Cultural History of the English Language This page intentionally left blank A Cultural History of the English Lan...

A History of the English Language

A History of the English Language Fifth Edition Baugh and Cable’s A History of the English Language has long been consi...

Towards a History of the Basque Language

Oops, page PA180 was not yet downloaded :( Oops, page PA181 was not yet downloaded :( Oops, page PA183 was not yet ...